{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# SOA 2019 Predictive Modeling, Innovation and Industry Contest \n", "# Computer Vision in Healthcare\n", "\n", "## Executive Summary\n", "\n", "\n", "The proposal for the Society of Actuaries (SOA) 2019 Predictive Modeling, Innovation and Industry Contest, is aimed to illustrate the potential of computer vision algorithms in healthcare. The primary objective is to design computer vision algorithms that improve diagnosis accuracy by reducing false positives and false negatives. The current proposal studies the HAM10000, Human Against Machine 10000, a dermatostopic image database [1]. The objective of the algorithm is to classify images of seven skin conditions, malignant, and non malignant, which include: Actinic Keratoses, Basal cell carcinoma, Benign keratosis, Dermatofibroma, Melanocytic nevi, Melanoma, and Vascular skin lesions. The algorithm presented is based on a Convolutional Neural Network (CNN), with ResNet50 [2] backbone, pre-trained on Imagenet weights and fine tuned with the current data. The winners of the International Skin Imaging Collaboration (ISIC) 2018 challlenge achieved a balanced multiclass accuracy of 88.5% in the competition test set. Unfortunately, the ISIC 2018 leaderboard is not accepting new submissions so we can not replicate the score on the test set. However, we replicated the 5-fold cross-validation score as reported by the winners of the challenge. Our algorithm achieves 67.9% balanced multiclass accuracy using 5-fold cross-validation on ResNet50, using only the 10,015 training samples. This result is similar to the 72.1% to 77.0% 5 fold cross-validation scores of ResNet-50 reported by the competition winners, that used 47,822 training samples. Our proposed algorithm uses only one neural network trained on ResNet50. The winners of the ISIC 2018 challenge used an ensemble of more than five models and external data, which is computationally more expensive [3]. \n", "\n", "\n", "\n", "This artificial intelligence tool is aimed primarily for primary care physicians, as they refer to specialist level of care. With a more accurate diagnosis, the use of dermatological care and biopsies may be reduced, reducing overall healthcare costs. The tool could also be used by health insurance companies to validate their diagnosis, aid in the underwritting process and evaluate providers in terms of diagnosis accuracy [4]. This tool is not intended to replace the dermatologist judgment, as an underlying condition may have different risk factors (such as age, gender, progression of the lesion, etc.) that are not captured by these artificial intelligence (AI) tools based on image data only. \n", "\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Introduction" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "\n", "Applications of computer vision in healthcare are numerous. Several examples include: X-ray pneumonia detection [5], breast cancer classification [6], skin cancer detection, [7] among others. An example of a start up that leverages this technology is entilic https://www.enlitic.com/. Entilic specializes in helping radiologists interpret X-rays, CT scans and MRI studies, in order to prioritize their work. Entilic claims that it has discovered malignant tumor nodules up to 18 months faster than standard of care.\n", "\n", "An SOA report (Shang, 2018) highlights the importance of computer vision in insurance applications [4]. In this article, the author presents a solution to a distracted driver problem, using CNNs, and also describes the potential of computer vision in other insurance areas, such as underwritting and healthcare. Our proposal focuses on providing more details of the implementation of computer vision techniques in healthcare. In particular, our proposal focuses on skin cancer detection, but the analysis can be extrapolated to other health conditions. As noted by Shang (2018), medical images can provide richer information for health insurance underwriting and pricing.\n", "\n", "The jupyter notebook is divided in several sections. First, we will present a brief overview of the technical requirements to run computer vision models. Second, we will provide a descriptive analysis of the HAM10000 dataset. After that we will cover the model set up, and the creation of the training and validation sets using cross-validation. The next step will illustrate the CNN architecture ResNet50, and the selection of the model parameters. A final step will describe the analysis of the results, and the limitations of the analysis. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Problem Definition and Risks\n", "\n", "\n", "Computer vision has a lot of applications in the health insurance marketplace. As noted by Shang, 2018, health insurance is an area where image recognition can provide more insight regarding the risk of individual customers. Physicians can use image recognition for diagnosis and prognosis [4]. In addition to linking doctors’ opinions to insurance pricing, given enough image data and claim experience, medical images such as computed tomography (CT) images of cancer patients may be used directly as pricing factors to help underwriting and derive risk loading and insurance premium rate [4]. \n", "\n", "\n", "Shang, 2018, also notes the risk of the computer vision models. Regulatory and reputational risk may also occur with the application of image recognition [4]. Image data may not be considered private and thus may be judged inappropriate for determining insurance rate and claim [4]. If an insurance company does not fully explain its methods to customers, regulators and the public, using nonpublic image data could cause reputational incidents [4]. Even if the data are usable, the way they are used to assess risks and determine insurance premium may be questioned [4]. Clients may want to know exactly how the algorithm determines the risk loading for insurance products [4]. Insurers that want to use image recognition technology need to be prepared for explanation and communication [4]. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Technical Details" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "This notebook was created using Jupyter notebook version 5.6.0, and Python 3.7.0. The operating sytem used is Ubuntu 18.04, on a computer using an NVIDIA 2080 GPU card (8GB memory size). For image processing, the use of a solid state hard drive is recommended to speed the training process due to the number of read and write operations involved in the training of computer vision models.\n", "\n", "This notebook uses a Graphic Processing Unit NVIDIA GPU. Configuration instructions for the deep learning environment are available on the NVIDIA website https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html. \n", "\n", "The Python libraries used for the analysis include: fastai (version 1.0.56.dev0), a machine learning library that includes code for computer vision models, pandas (version 0.24.2), a library for data manipulation, sklearn (version 0.20.2) for selection of the training, validation, and test sets and for calculation of important metrics, cv2 (version 3.4.3), for image processing, and matplotlib (version 3.0.0), for general purpose graphics." ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "fastai version 1.0.56.dev0\n", "Pandas version 0.24.2\n", "sklearn version 0.20.2\n", "cv2 3.4.3\n", "matplotlib 3.0.0\n" ] } ], "source": [ "%reload_ext autoreload\n", "%autoreload 2\n", "%matplotlib inline\n", "\n", "import matplotlib.pyplot as plt\n", "\n", "#Create symbolic link for fastai path - only once\n", "#! ln -s /home/username/fastai/fastai\n", "\n", "\n", "#Load libraries\n", "import fastai\n", "import sklearn\n", "\n", "from fastai.vision import *\n", "import pandas as pd # For data manipulation\n", "\n", "from fastai.callbacks import *\n", "\n", "\n", "from sklearn.model_selection import train_test_split\n", "\n", "\n", "\n", "import cv2\n", "import matplotlib \n", "\n", "print (\"fastai version\", fastai.__version__)\n", "print (\"Pandas version\", pd.__version__)\n", "print (\"sklearn version\", sklearn.__version__)\n", "\n", "print (\"cv2\", cv2.__version__)\n", "print (\"matplotlib\", matplotlib.__version__)\n", "\n", "\n", "#To handle LaTeX references\n", "#https://github.com/Jupyter-contrib/jupyter_nbextensions_configurator\n", "#!pip install jupyter_nbextensions_configurator\n", "#jupyter nbextensions_configurator enable --user\n", "#!pip install jupyter_contrib_nbextensions\n", "#!jupyter contrib nbextension install --user" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Data Analysis\n", "\n", "The HAM10000 database (Human Against Machine with 10000 training images), contains 10015 dermatoscopic images which are released as a training set for academic machine learning purposes [1]. The database is publicly available at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/DBW86T, and it is released under a Creative Commons Attribution-NonCommercial 4.0 International Public License.\n", "\n", "The HAM10000 database database contains 10,015 images of skin conditions corresponding to 7,470 unique lesions, whith some lesions having as many as six pictures. We assume that each lesion corresponds to an independent patient. The database includes information for seven diagnosis including: Actinic Keratoses (Solar Keratoses) and Intraepithelial Carcinoma (Bowen’s disease) (akiec) (N=228), Basal cell carcinoma (bcc) (N=327), Benign keratosis (bkl) (N=727), Dermatofibroma (df) (N=73), Melanocytic nevi (nv) (N=5,403), Melanoma (mel) (N=614), and Vascular skin lesions (vasc) (N=98). The average age of the patients in the database is 52 years (standard deviation 16 years) (N = 7,418; 0.70 % missing data) with a minimumm age of 0 months and maximum age of 85 years, and 54% male gender (N = 4,001). The images of the lessions come from fifteen diferent regions, such as lower extremity 21% (N=1,584), back 21% (N= 1,534), trunk 17%(N= 1,269), abdomen 11% (N=822), and others 30% (N=2,261). \n", "\n", "As can be seen from the frequency counts of the skin conditions, we have an unbalanced classification problem, with the most prevalent diagnosis being Melanocytic nevi (N = 5,403 patients) and the least prevalent diagnosis being Dermatofibroma (N=73 patients)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Data Quality\n", "\n", "Data was reviewed for consistency according to the Actuarial Standards of Practice (ASOP) 23. The metadata is believed to be adequate for the purposes of this analysis, with a small percentage of missing data in the age field (0.70 % missing data N = 52 patients), and we didn't identify any potential outliers. For the image data, Tschandl, et. al [1] state that they perform manual changes to image histograms to correct for over/under exposure and undesiredcolour shifts. Aditionally, Tschandl, et al [1] also state that more than 50% of lesions have been confirmed by pathology, while the ground truth for the rest of the cases was either follow-up, expert consensus, or confirmation by in-vivo confocal microscopy. Therefore, the data is deemed apropriate for the purposes of this analysis. However, material changes in the data could potentially yield to significant changes in the inference." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Load Metadata " ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "#Load metadata\n", "metadata = pd.read_csv('/home/user/Documents/HAM_Dataset/HAM10000_metadata.csv')" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [], "source": [ "#Path to the folder containing the images\n", " \n", "path_img = '/home/user/Documents/HAM_Dataset/Data/'\n" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
lesion_idimage_iddxdx_typeagesexlocalization
0HAM_0000118ISIC_0027419bklhisto80.0malescalp
1HAM_0000118ISIC_0025030bklhisto80.0malescalp
2HAM_0002730ISIC_0026769bklhisto80.0malescalp
3HAM_0002730ISIC_0025661bklhisto80.0malescalp
4HAM_0001466ISIC_0031633bklhisto75.0maleear
\n", "
" ], "text/plain": [ " lesion_id image_id dx dx_type age sex localization\n", "0 HAM_0000118 ISIC_0027419 bkl histo 80.0 male scalp\n", "1 HAM_0000118 ISIC_0025030 bkl histo 80.0 male scalp\n", "2 HAM_0002730 ISIC_0026769 bkl histo 80.0 male scalp\n", "3 HAM_0002730 ISIC_0025661 bkl histo 80.0 male scalp\n", "4 HAM_0001466 ISIC_0031633 bkl histo 75.0 male ear" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "metadata.head()" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "patient_data = metadata.drop_duplicates(['lesion_id'])" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Descriptive analysis of the Data" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
lesion_idimage_iddxdx_typeagesexlocalization
count100151001510015100159958.0000001001510015
unique74701001574NaN315
topHAM_0001863ISIC_0031764nvhistoNaNmaleback
freq6167055340NaN54062192
meanNaNNaNNaNNaN51.863828NaNNaN
stdNaNNaNNaNNaN16.968614NaNNaN
minNaNNaNNaNNaN0.000000NaNNaN
25%NaNNaNNaNNaN40.000000NaNNaN
50%NaNNaNNaNNaN50.000000NaNNaN
75%NaNNaNNaNNaN65.000000NaNNaN
maxNaNNaNNaNNaN85.000000NaNNaN
\n", "
" ], "text/plain": [ " lesion_id image_id dx dx_type age sex \\\n", "count 10015 10015 10015 10015 9958.000000 10015 \n", "unique 7470 10015 7 4 NaN 3 \n", "top HAM_0001863 ISIC_0031764 nv histo NaN male \n", "freq 6 1 6705 5340 NaN 5406 \n", "mean NaN NaN NaN NaN 51.863828 NaN \n", "std NaN NaN NaN NaN 16.968614 NaN \n", "min NaN NaN NaN NaN 0.000000 NaN \n", "25% NaN NaN NaN NaN 40.000000 NaN \n", "50% NaN NaN NaN NaN 50.000000 NaN \n", "75% NaN NaN NaN NaN 65.000000 NaN \n", "max NaN NaN NaN NaN 85.000000 NaN \n", "\n", " localization \n", "count 10015 \n", "unique 15 \n", "top back \n", "freq 2192 \n", "mean NaN \n", "std NaN \n", "min NaN \n", "25% NaN \n", "50% NaN \n", "75% NaN \n", "max NaN " ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "metadata.describe(include='all')" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
lesion_idimage_iddxdx_typeagesexlocalization
count74707470747074707418.00000074707470
unique7470747074NaN315
topHAM_0002673ISIC_0031764nvfollow_upNaNmalelower extremity
freq1154033704NaN40011584
meanNaNNaNNaNNaN51.658803NaNNaN
stdNaNNaNNaNNaN16.016472NaNNaN
minNaNNaNNaNNaN0.000000NaNNaN
25%NaNNaNNaNNaN40.000000NaNNaN
50%NaNNaNNaNNaN50.000000NaNNaN
75%NaNNaNNaNNaN65.000000NaNNaN
maxNaNNaNNaNNaN85.000000NaNNaN
\n", "
" ], "text/plain": [ " lesion_id image_id dx dx_type age sex \\\n", "count 7470 7470 7470 7470 7418.000000 7470 \n", "unique 7470 7470 7 4 NaN 3 \n", "top HAM_0002673 ISIC_0031764 nv follow_up NaN male \n", "freq 1 1 5403 3704 NaN 4001 \n", "mean NaN NaN NaN NaN 51.658803 NaN \n", "std NaN NaN NaN NaN 16.016472 NaN \n", "min NaN NaN NaN NaN 0.000000 NaN \n", "25% NaN NaN NaN NaN 40.000000 NaN \n", "50% NaN NaN NaN NaN 50.000000 NaN \n", "75% NaN NaN NaN NaN 65.000000 NaN \n", "max NaN NaN NaN NaN 85.000000 NaN \n", "\n", " localization \n", "count 7470 \n", "unique 15 \n", "top lower extremity \n", "freq 1584 \n", "mean NaN \n", "std NaN \n", "min NaN \n", "25% NaN \n", "50% NaN \n", "75% NaN \n", "max NaN " ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "patient_data.describe(include='all')" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "nv 5403\n", "bkl 727\n", "mel 614\n", "bcc 327\n", "akiec 228\n", "vasc 98\n", "df 73\n", "Name: dx, dtype: int64" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "patient_data['dx'].value_counts()" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "male 4001\n", "female 3419\n", "unknown 50\n", "Name: sex, dtype: int64" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" } ], "source": [ "patient_data['sex'].value_counts()" ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "lower extremity 1584\n", "back 1534\n", "trunk 1269\n", "abdomen 822\n", "upper extremity 776\n", "face 470\n", "chest 240\n", "foot 228\n", "unknown 203\n", "neck 119\n", "scalp 78\n", "hand 64\n", "genital 45\n", "ear 35\n", "acral 3\n", "Name: localization, dtype: int64" ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" } ], "source": [ "patient_data['localization'].value_counts() " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Selected Examples of the Skin Conditions \n", "\n", "Below we present an example image of each of the seven skin conditions studied in this database. The goal of the computer vision algorithm is to predict given an image, which one of the seven skin conditions the image corresponds to." ] }, { "cell_type": "code", "execution_count": 90, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Selected Examples of the Lessions\n", "#Set path \n", "path_img = '/home/user/Documents/HAM_Dataset/Data/'\n", "\n", "import warnings\n", "warnings.filterwarnings(\"ignore\")\n", "\n", "#function taken from https://www.datacamp.com/community/tutorials/face-detection-python-opencv\n", "def convertToRGB(image):\n", " return cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n", "\n", "fig = plt.figure(figsize=(10,10))\n", "\n", "#akiec ISIC_0029417\n", "\n", "ax1 = fig.add_subplot(331)\n", "ax1.set_title('Actinic Keratoses')\n", "ax1 = cv2.imread(path_img + 'ISIC_0029417.jpg')\n", "ax1 = plt.imshow(convertToRGB(ax1))\n", "\n", "#bcc ISIC_0028155\n", "\n", "ax2 = fig.add_subplot(332)\n", "ax2.set_title('Basal cell carcinoma')\n", "ax2 = cv2.imread(path_img + 'ISIC_0028155.jpg')\n", "ax2 = plt.imshow(convertToRGB(ax2))\n", "\n", "#bkl ISIC_0027419\n", "ax3 = fig.add_subplot(333)\n", "ax3.set_title('Benign keratosis')\n", "ax3 = cv2.imread(path_img + 'ISIC_0027419.jpg')\n", "ax3 = plt.imshow(convertToRGB(ax3))\n", "\n", "#df ISIC_0027008\n", "ax4 = fig.add_subplot(334)\n", "ax4.set_title('Dermatofibroma')\n", "ax4 = cv2.imread(path_img + 'ISIC_0027008.jpg')\n", "ax4 = plt.imshow(convertToRGB(ax4))\n", "\n", "#nv ISIC_0024698\n", "ax6 = fig.add_subplot(335)\n", "ax6.set_title('Melanocytic nevi')\n", "ax6 = cv2.imread(path_img + 'ISIC_0024698.jpg')\n", "ax6 = plt.imshow(convertToRGB(ax6))\n", "\n", "#mel ISIC_0025964\n", "ax5 = fig.add_subplot(336)\n", "ax5.set_title('Melanoma')\n", "ax5 = cv2.imread(path_img + 'ISIC_0025964.jpg')\n", "ax5 = plt.imshow(convertToRGB(ax5))\n", "\n", "#vasc ISIC_0031197\n", "ax7 = fig.add_subplot(337)\n", "ax7.set_title('Vascular skin lesions')\n", "ax7 = cv2.imread(path_img + 'ISIC_0031197.jpg')\n", "ax7 = plt.imshow(convertToRGB(ax7))\n", "\n", "fig.show()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Creation of the training and validation sets" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The next step in the modeling process is the creation of the training and validation sets. We do that using 5-fold stratified cross validation to get folds that have similar proportions of every diagnosis. Since there are several lesions with repeated pictures, we make sure that no lesions overlap between the train and validation sets. Doing so is called a validation set leakage, and will most likely cause overfitting, as the model will learn to memorize the images of a particular condition and not generalize well in unseen data." ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [], "source": [ "#Creation of a database by lession id\n", "\n", "from sklearn.model_selection import StratifiedKFold\n", "skf = StratifiedKFold(n_splits=5, random_state=42)\n", "\n", "lesion_id = pd.DataFrame({'lesion_id':list(metadata['lesion_id'].value_counts().index)})\n", "\n", "patient_data = metadata.drop_duplicates(['lesion_id'])\n", "patient_data = patient_data.reset_index(drop=True)\n", "X = patient_data['lesion_id']\n", "y = patient_data['dx']\n" ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [], "source": [ "train_1, train_2, train_3, train_4, train_5 = [patient_data.loc[i[0]] for i in skf.split(X, y)]" ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [], "source": [ "valid_1, valid_2, valid_3, valid_4, valid_5 = [patient_data.loc[i[1]] for i in skf.split(X, y)]" ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [], "source": [ "train_1 = train_1.assign(database = \"train\")\n", "train_1 = train_1.assign(is_validation = False)\n", "valid_1 = valid_1.assign(database = \"validation\")\n", "valid_1 = valid_1.assign(is_validation = True)" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [], "source": [ "train_2 = train_2.assign(database = \"train\")\n", "train_2 = train_2.assign(is_validation = False)\n", "valid_2 = valid_2.assign(database = \"validation\")\n", "valid_2 = valid_2.assign(is_validation = True)\n", "\n", "train_3 = train_3.assign(database = \"train\")\n", "train_3 = train_3.assign(is_validation = False)\n", "valid_3 = valid_3.assign(database = \"validation\")\n", "valid_3 = valid_3.assign(is_validation = True)\n", "\n", "train_4 = train_4.assign(database = \"train\")\n", "train_4 = train_4.assign(is_validation = False)\n", "valid_4 = valid_4.assign(database = \"validation\")\n", "valid_4 = valid_4.assign(is_validation = True)\n", "\n", "train_5 = train_5.assign(database = \"train\")\n", "train_5 = train_5.assign(is_validation = False)\n", "valid_5 = valid_5.assign(database = \"validation\")\n", "valid_5 = valid_5.assign(is_validation = True)" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [], "source": [ "#Concatenate the databases\n", "metadata_train_valid = pd.concat([train_1, valid_1]) \n", "metadata_train_valid2 = pd.concat([train_2, valid_2])\n", "metadata_train_valid3 = pd.concat([train_3, valid_3])\n", "metadata_train_valid4 = pd.concat([train_4, valid_4])\n", "metadata_train_valid5 = pd.concat([train_5, valid_5])\n" ] }, { "cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [], "source": [ "metadata_train_valid = pd.merge(metadata, metadata_train_valid[['lesion_id', 'is_validation']], on=['lesion_id'])\n", "metadata_train_valid2 = pd.merge(metadata, metadata_train_valid2[['lesion_id', 'is_validation']], on=['lesion_id'])\n", "metadata_train_valid3 = pd.merge(metadata, metadata_train_valid3[['lesion_id', 'is_validation']], on=['lesion_id'])\n", "metadata_train_valid4 = pd.merge(metadata, metadata_train_valid4[['lesion_id', 'is_validation']], on=['lesion_id'])\n", "metadata_train_valid5 = pd.merge(metadata, metadata_train_valid5[['lesion_id', 'is_validation']], on=['lesion_id'])" ] }, { "cell_type": "code", "execution_count": 19, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
is_validationFalseTrue
dx
akiec26859
bcc400114
bkl810289
df8827
mel890223
nv56231082
vasc10537
\n", "
" ], "text/plain": [ "is_validation False True \n", "dx \n", "akiec 268 59\n", "bcc 400 114\n", "bkl 810 289\n", "df 88 27\n", "mel 890 223\n", "nv 5623 1082\n", "vasc 105 37" ] }, "execution_count": 19, "metadata": {}, "output_type": "execute_result" } ], "source": [ "pd.crosstab(metadata_train_valid.dx, metadata_train_valid.is_validation) #metadata_train_valid['dx'].value_counts()" ] }, { "cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
is_validationFalseTrue
dx
akiec26265
bcc396118
bkl914185
df8530
mel847266
nv56241081
vasc9943
\n", "
" ], "text/plain": [ "is_validation False True \n", "dx \n", "akiec 262 65\n", "bcc 396 118\n", "bkl 914 185\n", "df 85 30\n", "mel 847 266\n", "nv 5624 1081\n", "vasc 99 43" ] }, "execution_count": 20, "metadata": {}, "output_type": "execute_result" } ], "source": [ "pd.crosstab(metadata_train_valid2.dx, metadata_train_valid2.is_validation) #metadata_train_valid2['dx'].value_counts()" ] }, { "cell_type": "code", "execution_count": 21, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
is_validationFalseTrue
dx
akiec25671
bcc43579
bkl886213
df8530
mel857256
nv56241081
vasc11824
\n", "
" ], "text/plain": [ "is_validation False True \n", "dx \n", "akiec 256 71\n", "bcc 435 79\n", "bkl 886 213\n", "df 85 30\n", "mel 857 256\n", "nv 5624 1081\n", "vasc 118 24" ] }, "execution_count": 21, "metadata": {}, "output_type": "execute_result" } ], "source": [ "pd.crosstab(metadata_train_valid3.dx, metadata_train_valid3.is_validation) #metadata_train_valid3['dx'].value_counts()" ] }, { "cell_type": "code", "execution_count": 22, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
is_validationFalseTrue
dx
akiec25176
bcc412102
bkl847252
df10114
mel942171
nv48821823
vasc12319
\n", "
" ], "text/plain": [ "is_validation False True \n", "dx \n", "akiec 251 76\n", "bcc 412 102\n", "bkl 847 252\n", "df 101 14\n", "mel 942 171\n", "nv 4882 1823\n", "vasc 123 19" ] }, "execution_count": 22, "metadata": {}, "output_type": "execute_result" } ], "source": [ "pd.crosstab(metadata_train_valid4.dx, metadata_train_valid4.is_validation) # metadata_train_valid4['dx'].value_counts()" ] }, { "cell_type": "code", "execution_count": 23, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
is_validationFalseTrue
dx
akiec27156
bcc413101
bkl939160
df10114
mel916197
nv50671638
vasc12319
\n", "
" ], "text/plain": [ "is_validation False True \n", "dx \n", "akiec 271 56\n", "bcc 413 101\n", "bkl 939 160\n", "df 101 14\n", "mel 916 197\n", "nv 5067 1638\n", "vasc 123 19" ] }, "execution_count": 23, "metadata": {}, "output_type": "execute_result" } ], "source": [ "pd.crosstab(metadata_train_valid5.dx, metadata_train_valid5.is_validation) #metadata_train_valid5['dx'].value_counts()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Modeling " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The training takes place in the fastai library [8], which sits on top of a popular AI library Pytorch. Installation instructions for the fastai library can be found at https://docs.fast.ai/install.html. \n", "\n", "The architecture selected for the CNN is ResNet50 [2], pre-trained on imagenet weights. ResNet is a popular architecture for training deep neural networks that uses a residual learning framework [2]. We initialize the network weights using Imagenet. Training takes place in two stages, first with image size to 224 * 224, we train only the heads (this is the last part of the network), then we unfreeze the other layers and train the rest of the network. Original image size is 600 * 450.\n", "\n", "To train the neural network we need to normalize the data, to do so, we calculate the statistics of the images mean and standard deviation, and use this to normalize the image set." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Model parameters\n", "\n", "Drop out regularization, a method to remove connections to reduce overfitting, was set to 50%.\n", "\n", "Batch size: This is the number of images that get processed by the learning process at a time. This depends on the GPU memory. In this case, using an NVIDIA 2080 card we can accomodate 32 images of size 224 * 224. \n", "\n", "Loss Function: We optimize the loss function cross entropy loss with weights. To deal with class imbalance we use the inverse of the frequency of each one of the seven skin conditions. This is, the loss function gets penalized more heavily if the prediction of a rare class is wrong. If we don't use a penalized loss function, the network tends to predict the most frequent class.\n", "\n", "Learning rate: After we create the weights, we also need to find the learning rate. This is one of the most important parameters when training neural networks. We use this by using cyclical learning rates as described in https://arxiv.org/abs/1506.01186.\n", "\n", "Data augmentation is also a technique to artificially create new data and reduce overfitting. This is availabe in fastai, with the function tfms = get_transforms(flip_vert = True). The transformations applied to the data are: random flip, flip vertical, rotation, zoom, a random lightning and contrast change, and a random symmetric warp.\n", "\n", "Test time augmentation was used at the end of every fold to improve the accuracy of the predictions. This is accomplished by taking every image, applying transformations, and taking the average of all of the predictions. " ] }, { "cell_type": "code", "execution_count": 24, "metadata": {}, "outputs": [], "source": [ "#Make the training process deterministic \n", "#function taken from kernel https://www.kaggle.com/iafoss/hypercolumns-pneumothorax-fastai-0-831-lb\n", "\n", "SEED = 2019\n", "\n", "def seed_everything(seed):\n", " random.seed(seed)\n", " os.environ['PYTHONHASHSEED'] = str(seed)\n", " np.random.seed(seed)\n", " torch.manual_seed(seed)\n", " torch.cuda.manual_seed(seed)\n", " torch.backends.cudnn.deterministic = True\n", " #tf.set_random_seed(seed)\n", "seed_everything(SEED)\n" ] }, { "cell_type": "code", "execution_count": 25, "metadata": {}, "outputs": [], "source": [ "#Normalization statistics \n", "import cv2\n", "\n", "\n", "img_stats = []\n", "\n", "def open_image(fname):\n", " img = cv2.imread(os.path.join(path_img, fname + '.jpg'))/255\n", " #img = cv2.resize(img, (sz, sz)) / 255\n", " x_tot = img.mean() #image statistics\n", " x2_tot = (img**2).mean()\n", " \n", " return x_tot, x2_tot\n", "\n", "\n", "for fname in metadata['image_id']:\n", " img_stats.append(open_image(fname))\n" ] }, { "cell_type": "code", "execution_count": 26, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "mean: 0.6267264561492303 , std: 0.18320752802569892\n" ] } ], "source": [ "\n", "#function taken from https://www.kaggle.com/iafoss/data-repack-and-image-statistics\n", "def get_stats(stats): # get dataset statistics \n", " x_tot, x2_tot = 0.0, 0.0\n", " for x, x2 in stats:\n", " x_tot += x\n", " x2_tot += x2\n", " \n", " img_avr = x_tot/len(stats)\n", " img_std = np.sqrt(x2_tot/len(stats) - img_avr**2)\n", " print('mean:',img_avr, ', std:', img_std)\n", " \n", " \n", "get_stats(img_stats)\n", "\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold 1" ] }, { "cell_type": "code", "execution_count": 28, "metadata": {}, "outputs": [], "source": [ "epochs_head = 4\n", "epochs_all = 15\n", "patience = 3" ] }, { "cell_type": "code", "execution_count": 29, "metadata": {}, "outputs": [], "source": [ "from sklearn.metrics import confusion_matrix\n" ] }, { "cell_type": "code", "execution_count": 30, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[30.62691131498471, 19.48443579766537, 9.112829845313922, 87.08695652173913, 8.998203054806828, 1.493661446681581, 70.52816901408451]\n" ] } ], "source": [ "#Class weights to deal with class imbalance\n", "#New weights - inverse of class frequency\n", "\n", "sample_size = 10015\n", "num_classes = 7\n", "\n", "\n", "w1 = sample_size/(327)\n", "w2 = sample_size/(514)\n", "w3 = sample_size/(1099)\n", "w4 = sample_size/(115)\n", "w5 = sample_size/(1113)\n", "w6 = sample_size/(6705) \n", "w7 = sample_size/(142)\n", "\n", "\n", "weights = [w1, w2, w3, w4, w5, w6, w7]\n", "class_weights = torch.FloatTensor(weights).cuda()\n", "\n", "\n", "print(weights)" ] }, { "cell_type": "code", "execution_count": 31, "metadata": {}, "outputs": [], "source": [ "#Increase image size\n", "\n", "\n", "gc.collect()\n", "\n", "bs = 32 #batch size - depends on the GPU memory and number of GPUs available. For The current GPU we have 2080 with\n", "#memory 8GB, we can use bs =16\n", "\n", "\n", "def get_data(metadata_train_valid, bs, size):\n", " stats = ([0.626, 0.626, 0.626], [0.1829, 0.1829, 0.1829])\n", " \n", " data = (ImageList.from_df(metadata_train_valid, path_img, cols='image_id', suffix='.jpg')\n", " .split_from_df(col='is_validation')\n", " .label_from_df(cols='dx')\n", " .transform(size=(size, size), tfms = get_transforms(flip_vert = True))\n", " .databunch(bs=bs)).normalize(stats)\n", " \n", " return data\n", "\n", "\n", "data = get_data(metadata_train_valid, bs, 224)\n", "\n", "#Define Convolutional Neural Network with backbone resnet50\n", "learn = cnn_learner(data, models.resnet50, metrics = [error_rate, accuracy],\n", " callback_fns=[ShowGraph])\n", "\n", "learn.loss_func=nn.CrossEntropyLoss(weight=class_weights)\n" ] }, { "cell_type": "code", "execution_count": 32, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Learn rate finder\n", "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 33, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 02:57

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
01.4955041.2196840.4265430.57345700:43
10.9393200.8889960.3200440.67995600:45
20.8161920.9459540.3440740.65592600:43
30.7040940.7876980.2878210.71217900:44
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.5734571218490601.\n", "Better model found at epoch 1 with accuracy value: 0.679956316947937.\n", "Better model found at epoch 3 with accuracy value: 0.7121791243553162.\n" ] } ], "source": [ "#Fit heads\n", "seed_everything(SEED)\n", "\n", "learn.clip_grad(1.0);\n", "\n", "lr = 1e-3\n", "learn.fit_one_cycle(epochs_head, max_lr = lr, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_resnet50_fold1')])" ] }, { "cell_type": "code", "execution_count": 34, "metadata": {}, "outputs": [], "source": [ "learn.load('Best_model_sz224_resnet50_fold1')\n", "learn.unfreeze()" ] }, { "cell_type": "code", "execution_count": 35, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 36, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 14:48

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
00.6302040.8397540.3014750.69852500:59
10.6410560.8343590.3014750.69852500:59
20.6911750.8040680.2900050.70999500:59
30.5993630.7847510.2834520.71654800:58
40.6231510.8029240.2867290.71327100:58
50.6030480.8364520.2992900.70071000:58
60.6174170.8473940.2960130.70398700:58
70.5178020.7944180.2856360.71436400:58
80.5398480.7910490.2807210.71927900:58
90.5226660.7998930.2850900.71491000:58
100.4989640.7689120.2747130.72528700:58
110.5292190.7537890.2637900.73621000:58
120.4913200.7818690.2758060.72419400:58
130.5077040.7743080.2752590.72474100:58
140.5212560.8089080.2889130.71108700:58
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvqOYd8AAAIABJREFUeJztnXeYVNXZwH/vdpZdYIGll6X33ixIsKAUxa7YjUZjYknUxGCixFgiJpYvKmrU2GKMMRgjERBFwYIgLEV6731Z6i7L1vP9ce/M3Jm5UxZ2d5bh/T3PPHvvuWfufefuzHvPedsRYwyKoihKfJEQawEURVGUqkeVu6IoShyiyl1RFCUOUeWuKIoSh6hyVxRFiUNUuSuKosQhqtwVRVHiEFXuiqIocYgqd0VRlDgkKVYXbty4scnJyYnV5ZV4wFRAeQmUlVh/y4v99yvK/Psnp0OD1tZfRTlJWbhw4T5jTHakfjFT7jk5OeTm5lbtSfetg/VfQFp991dKBiToZOWkobQIDm61X1t82wfs7aP7/PsnpkKDHGjQFhq0cbzawv6N8NlDULgTBv0EznkI6jSIycdSlBNBRLZE0y9myr1a2L4APv1N6OOSAKn1bGVfD9IahH4QeF6e/nXsvkr1UJAHyz6A7bk+JV64179PYgrUb20p7K5jrL9ZOT4lXrdJ6Id360HQZSR8+QQseA1W/hfOfwJ6XwUi1f7xFKWmkVgVDhs4cKCp8pF7eRkcOwTHDlp/iw/b+4GvEO0lR8Kfv0kP6HU59LzcUirKiVFeBhu+gMV/hzXTLTNKg7YOhd0Wshyj8IxmVTPz2vUDfHIf7MiFtkNhzDPQpOuJn1dRagARWWiMGRixX1wp9xOlojz0A6FgL6z9FLZ9b/VtNRh6XQE9LoWMJrGV+2QjfwMs+QcseQ+O7IL0xtBnHPS7Hpp0qxkZKipg8Tvw+e+hpABOvxOGPQCpGTVzfUU5TlS5VxcHtsDyD63XnuWWqafdMOh5BXS7SO24oSg5CqumwKK/w5ZvrfvWcYSl0DuPhKSU2MhVuA9m/h4Wvwv1WsGoidD1QjXV1FJKS0vZvn07x44di7Uo1U5aWhqtWrUiOTnZr12Ve02wdzUsnwzLJsOBTZZNuNP5ltmm80hIOcWjMoyBnYsshb78Q2tWlNXOUuh9r4V6LWItoY+t8yxTzd4V1v9w1J+gYbtYS6UEsGnTJjIzM2nUqBESxw9gYwz5+fkcOXKEdu38v4eq3GsSY2DHIkvRL/8PFOy2InO6jLZMNx3OgcTkyOeJFwrzYem/rNHw3hWQVAe6Xwz9b4C2Z9beUXF5Gcz/K8z6o2X/P+t+OOMeSE6LtWSKzapVq+jatWtcK3YPxhhWr15Nt27+pspolXt8RcvEChFoNcB6nf84bJljjeZXfmxFgNTJspRbzyss5RaP4ZgV5bBhluUcXT0VKkqhRX+48DlrJnMyRBolJlm29x6Xwozfwawn4If3YfSfoeO5sZZOsTkVFDuc+OeMSrmLyEjgL0Ai8LoxZmLA8TbA20ADu894Y8y0E5LsZCUh0bLBtxsGo5+2okGWTYalH8DCtyCzBfS8zFJ4LfrV3lFstOzf5HOOHt4BdRrC4Nss00vTHrGW7vio1wKufNOaaUz9Fbx7GXS/BEY+WbtMScqpRdEBS49ESUSzjIgkAmuBEcB2YAFwjTFmpaPPq8BiY8zLItIdmGaMyQl33rgyy0RDSaEV7rdsMqyfaY1sG7a3RvNdx0DTntbI8WSgtAhW/c8apW/6GhBrZNvvessUlZQaawmrjrJimPM8fPM0JCTB8AdhyE9PLTNbLWLVqlVBZoqa5ODBg7z33nv8/Oc/r9T7Ro8ezXvvvUeDBpULuFi1ahXd6h6yBoYrPoKyY8gfDleNzV1ETgceMcZcYO8/CGCMedLR56/ARmPMU3b/Z4wxZ4Q77ymn3J0c3W8px+WTYdM3gIHkutCyP7QaZL1aD4a6jWMtqeVPOLgVdiy0nKM7FsPOxVBaaMWhe5yj9VvFWtLqZf8mmP4ArPvMyne48Floc1qspaocpccsf5AxJ62zONbKffPmzVx44YUsX77cr72srIykpCocnJWXQdF+Vq1cTrdpl0JKppVwN+AmpEXfKrO5twS2Ofa3A0MC+jwCfCYidwN1gfOi/AinJukNYcBN1uvIbkvBb58P2+bDd8/7aqJktbOUvEfZN+lR/aP7wn2Wc9irzBfC0XzrWGIqNOsF/a6zwgVzzopP/4EbDdvBtR9Y/oTpv4E3LoC+18OIP8T+IVxeCgV7rO/SkV2h/xYd8L2naS/ofaU1c6zfMnayn2SMHz+eDRs20LdvX5KTk0lLSyMrK4vVq1ezdu1aLrnkErZt28axY8f4xS9+we233w74yq0UFBQwatQohg4dynfffUfLli35+OOPqVOnjvXQLSm0ymoUHQQMkABjX7RMuSl1KyVrNCP3K4CRxpif2Ps3AEOMMXc5+txnn+sZe+T+N6CnMaYi4Fy3A7cDtGnTZsCWLVGVSDi1KDkKu5ZYin77AuuvJw3fO7ofaCVRnejovrjAupZTmR/cah8UK6GoRX/rmi37Ww+XWMWj1yZKCuHrP8N3L1g/uOxukFzHKkiWXCfgle5yLLDN/puUZm0npVq+mIpy62F7ZJfj5aK4C/dhKQIHkgiZzexXc//t4iOWeXBHLiCQMxR6XWk5/Wt5noZz5P6H/61g5c7DVXr+7i3q8fuLQvuKnCP32bNnM2bMGJYvX+4NV9y/fz8NGzakqKiIQYMG8dVXX9GoUSM/5d6xY0dyc3Pp27cvV111FWMvHMP1l420oszKi63/XXoWpDdm1frN1RotswNo7dhvZbc5uRUYCWCMmSsiaUBjwK84iDHmVeBVsMwyUVz71CMlHdqeYb3ANotsgW0LLGW/fb6lVCo7ui8rscISdyy0TCs7FsK+NVZlRbDS+1sOgEG3WX+b99FszVCk1IXzHoHe4+CbZ6xRc0mhpWRLj1o+Cc/f8uLjuIBYSr7sGJjy4GMZTSxFXa+l9b/yKm/H3/RGlnM/FKf9zMoUXjbZiuj63z0w7VfQ+QLodZUV668hoBEZPHiwXxz6888/z0cffQTAtm3bWLduHY0aNfJ7T7t27ejbpw8UH2FAt/ZsXj4fzuttDd4ym1o1r8L976IkGuW+AOgkIu2wlPo44NqAPluBc4G3RKQbkAbknbB0ijWCy8qxXr2vtNoCR/cbZllx5WAphRb9rUJZWe2sLNodi2D3Mp+iSW9sjcR7XOIbmcfatHAy0qQrXP5a+D4V5bayL4KyIn/F7/0b2HbM+puUBvWa+yvtuk2qzjTXqAMM/w386AFr1rb031ay2ar/QWp96D7WsvO2HVorzW/hRtg1Rd26PlPJ7NmzmTlzJnPnziU9PZ3hw4cHZ9KWl5KanAh7V0F5MYmmlKKEVMjuas3eqpCI3xJjTJmI3AXMwApzfMMYs0JEHgVyjTFTgPuB10TkXqz54c0mVtlRpwKuo/utPjOOc3SfXNcKuRxyuzXKa9HfGqWf7CGYJwsJidYMqDbPgkSs70ZLO09j01ew7N9WdMbiv1szhJ6XW4q+ac/4+O6Ul0HxIetBmpBoRUIlJlt/va/g0XNmZiZHjrgXGDx06BBZWVmkp6ezevVq5s2bZx0wxnod2AL791i/y8Qkn6mssLDKFTtEGedux6xPC2ib4NheCZxZtaIpUSNiVU/MamtlxII1uj+yyxrxV8EUTzlFSEyywlo7ngtjnoU10yxFP+8ly9mf3c2aQfa60hoknEyUl1oVY4sOWsXiABCC/BUeJMGh6JMhMYlGyUmcOWQAPXt0o06dOjRt2tQ6b0ISI0eO5JVXXqFbt2506dKF04YMsSLj9q60Qp+LC6yExsRUaNzZvkb1PSi1/ICiKJEpzIeVH1mmm232iLTNGZai736JFQFWA1Q6FLKs2KrqWnTQCt8FS7nWaWDZtpPrWH6nijLfq7zMUsZ++462UDhH/SKWMsdYpUjSG9m29MqZt9w+r5YfUBSl6qjbyFrBatBP4MBmX9b1J/fCtAeg0whr1tj6NCuLN5amm9Jj9poOBy0/Blj1jTKbW2UwktL85ZNEe3YbRfKdMZYfJUj5uzwM6mZbSj1GjmlV7oqiVI6sHBj2K6uw2u6llpJf/qFlwgFLgTbpbr+6WWUomnSzTBLVgTGWs7rIXqinzHZiJqdbD5q0BlWXNS1ima5Ogmzy2i+hoii1ExErZLZ5HxjxqOXQ373MsjHvWWmN7osP+fpntrCVfXcrZLdJN8jucnzORGOsiKIie4ReXmK1p2RYdfnT6p/yORmq3BVFOXESEq1yDM6SDMZYxeT2roI9K6y/e1fA99/6wnIlwaqx5BnpN7X/NmwfHAhQUW6Nyg9tt5R6RSkgViRSRlNLoWvNHy+q3BVFqR5ErJpD9VtZNnkP5WWwf6M1wve89qyw4us9kStJaVZESdMeVgz4/o1W6YezXoJCsRaur9PAWug+QdWYG3pXFEWpWRKTILuz9epxia+95KiVNe0c6W+cDT/80zK3dDrfSrZr1kvDe6NAlbuiKLWDlHQr4a5FP//2ogNWtEtyGqxadVIp9oyMDAoKCti5cyf33HMPkydPDuozfPhwnn76aQYOjBjdWClUuSuKUrupriibGqRFixauir06qX0FI2KMVk1QFCUU48ePZ9KkSd79Rx55hMcff5xzzz2X/v3706tXLz7++OOg923evJmePXsCUFRUxLhx4+jWrRuXXnopRUVF1SKrjtwdPPXpal6evYGNfxxNQkIc1M9QlHhm+ngr9LIqadYLRk0Mefjqq6/ml7/8JXfeeScAH3zwATNmzOCee+6hXr167Nu3j9NOO42xY8eGXAP15ZdfJj09nVWrVrF06VL69+9ftZ/BRpW7g5dnbwBg75FimtXXcqeKovjTr18/9u7dy86dO8nLyyMrK4tmzZpx77338vXXX5OQkMCOHTvYs2cPzZo1cz3H119/zT333ANA79696d27d7XIGlPl/vXaPHq0qMeCzQe4492FLJkwggbpsU88uPnN+Xz6y2GxFkNRlHCEGWFXJ1deeSWTJ09m9+7dXH311fzjH/8gLy+PhQsXkpycTE5OTnCp3xgQM+VugBvfmE/XZplkpllirNl9hCHtG4V/Yw2werd7SU9FUZSrr76a2267jX379vHVV1/xwQcf0KRJE5KTk5k1axaRVpgbNmwY7733Hueccw7Lly9n6dKl1SJn7Byqtt9y7Z4jJCdaYpSWx86ZWVwWuOKNoihKMD169ODIkSO0bNmS5s2bc91115Gbm0uvXr1455136Nq1a9j3/+xnP6OgoIBu3boxYcIEBgwYUC1yxmzkfuRYKQAVBlKTLOVeUh47Bet8sHRtlhkzORRFqf0sW+Zz5DZu3Ji5c+e69isosOrG5+TksHz5cgDq1KnD+++/X+0yxmzkXlDsq4ucmmQlJRSXVoTqXu2UV/iUe1ryyZMkoSiK4kZUyl1ERorIGhFZLyLjXY4/JyJL7NdaETlYGSFSky0xjsXQNFLhUO7OB091Ul5heHrGGjbtK6yR6ymKcuoQ0SwjIonAJGAEsB1YICJT7KX1ADDG3OvofzfQL+hEYaiXZlVyyy8oqczbqpQKR/LSwaMlFJeVe2cU1cXGvAJenLWe5TsP8daPB1frtRQlXjDGhIwhjydONKEympH7YGC9MWajMaYEeB+4OEz/a4B/Rjqp85+TZo/cjxyrmRGzG+X2jUxNSmBfQQldHvqUv361oVqvWVxmmaHW7SmI0FNRFIC0tDTy8/PjPpPcGEN+fj5pacefbxONQ7UlsM2xvx0Y4tZRRNoC7YAvQxy/HbgdIKVZR5rb7WW2SWTGit0M75JNvzY1X0ti72GrvnRyYoJX6T45fTW3D2tfbaOEknLrOjsOFrF+bwGtsuqovV9RwtCqVSu2b99OXl5erEWpdtLS0mjVqtVxv7+qo2XGAZONMa7Gc2PMq8CrAKnNO3kfvaW2klu9+wiXvvQdmyeOqWKxIvN/M9cBwfb2oyXl1E2tnqAipwP5vGe/omuzTE2eUpQwJCcn065du1iLcVIQjVlmB9Dasd/KbnNjHFGYZAIpi2F8u4dQg/PqNBV5Ru4eVu8+wrFSjbdXFOXEiUa5LwA6iUg7EUnBUuBTAjuJSFcgC3AP+AxBneREdh+Ofapu75b1AbjlTP9RwQ/bKxX4UymKXRS5xzykKIpyIkRU7saYMuAuYAawCvjAGLNCRB4VkbGOruOA900lPR1FpeXMXhN7+5nH7t8yy3+x3p/+fWG1XTNw5B6qTVEUpbJEZUw2xkwDpgW0TQjYf6TqxIqOL1bt4eDRUi4fEJ3TYen2g3Rpluka4viXLyybe3ZmatCx6gq9KikLVuSezF2AY6XlzN2Qz/q9Bdw2rH2VX19RlPilVi7WEa3d+da3c7n/3z9ElXS061ARY1+cw0MfLQ865sxOvbBXcx67uAfTf3GWt63IRZ5jpeVMXrjdL/kpFKXlFSx1Me8Uuyj3S1/6zrt9/nNf8+O3FvDEtFX0mPBpxOsoiqJ4qJXK/a3vNleq/zWvzovYp6jEUtDfb9ofdOzT5bu92wkJwg2n59CmYbq3ze3h8f78rfzq3z/wwIe+im7Tlu3ivn8tCYrBfXn2Bsa+OIdFWw8AVvXLv327yXXk7qGwuIyt+4/69kvU0aooSvTUSuU+cfrqSvVftuMQ+wqK+d1Hy0JWd0xKsEscuIzCDztMIR6c4Y+FxcHv2V9oZdNOXrjd2/bzfyziP4t3BI3Iv9uwD4AHJlsPgstemsNjn6z0nuPSfi2Dzj912S7Xz6EoihINtUK5i0DjjOgX6Sgrr+DtgNH9wMdn8o/vtzLgsZnc/8EP3vYt+YVMX7aL0gpL4e49EhyNkmKXHL5mcBvX6xW4hEM2DVipqdAxul+/1z/j1DNb8LR7lL/Hzv/kZb145Xpf2c8DhSWu5p5v1+1zlU9RFCWQWqHcAQa2bei3v2BzsPnEw79yt/H7KStcjxUUl/HhIt9o+ppX5/GzfyzyLqEHloN0ybaDDPvTLF6wFSzAHT/yd1q+d9sQ7zkD8dTDATjjyS/43w87vfsXvvCtX9/A+KE6AVmoqUkJDOvc2Lu//UARdVKCnb7X/+17Dh711d+pqDCUaXSNoigu1ArlbkxwCOLvP3ZX3gAHjwabUdwoK69g5yErht5pPtlXUMKNf/uerfuP8szna1mzx1p5KTATNTvDipy57vV5vDVnk58t3VlobOehY0HvDWUeuuu9RRwJeFiIiJ/CLy4r55EQD6+xL85h075CDh0t5cY35tPxd9PdP7yiKKc0tUK5A9Svk8xp7X2j95W7Dofs6zGjAFwRJgxy3kb30f+SbQc57DC1vPr1RgCyAtZvbdHAeuBUGHjkfyvJ3XLAeyzQrn73Pxf77e84UOTdvqy/z6b+yVJ3W7qI8N87zwRg16FjHHA8wG53hEFu3X+Us5+eTZ9HP+Pb9ZaZJt6LKCmKUnlqjXLP3XKA928/nSYuceaBJCX6Ys4rwii26//2vWv7mt3uD47EBP9Y9kDzyepdh7lk0hx2HSryOkdDcc4zX3m3wxUDa5zh+7z17LVk9zgydr8bfw6/Hd0t7LUOFfnPZHLGTyVn/FT2FWi2q4c1u4+wKsyAQVHijVqj3BfYTsdoMjT/8D9vKXmu6O8+cs8ZPzXk+/85f1vIY04SApT9wx+vYMm2gzw9Y21U79+2/yjbDxxl8dbQJQy6NMvwbtevY9nxnU5fT9sr1/cPeY49IUoWPPPZmqjkjJbdh44x3yWU1ENhcVmt9QFc8H9fM+ov38RaDEWpMWqNcjf2itnOgfjzDmdnKM7o2Dhin0B2HCyK3CkMzpHyl/f/KGS/GSt2M/SpWd4R408dDttBOVZZ45eu9UXJ1LMV+Qe51sPnp8Pae235I7o3i0oeJ0erODZ+9PPfcNVf53JDiBlRj9/P4I53F1XpNRVFOT5qjXK3IxX5x098peKf/Tz8CPnhC7tHPG+zemmc0aFRxH5XDXSfAbx248CgNmesfFlAyGJdR5TL41NX+R07r1tT7/bfbx3C5oljqJ/ui7pJTkygbkqi12HcIds3qg80GTk54hKnD5BWxStJeeLyvwkTkjlz1Z4qvWZV89gnK8kZP1X9FErcU3uUu/1j69myvp8DMZBtjqzNW4f6V3Ac0DZ4kY/dh4/xzi2Rl7B78rLeru0jujcNauvYxKd0Ozm2ARY8dF7Ia2Q5FHkoO7zHiQvBDw4Pt53l/7kDyxJ7HjCtAiKQqpKXZ2/wZtyeTPzt202Ab3akKPFKrVHuVw70lYxPcBTpCrThOu3tHt77yRB+Oqw9bRulBx0DSEpMYFTPZrx6wwDX45snjgk7Mj69vf/I3xlrLiIMtU1Db948iPSUJMYNak0gVw9sTaYjNj4UTev5kqMCP8/gHCuaKNAUFThyT0my/q3VudD3U5+u5rKXvqu1NvZIuGUdK0o8UWuUu3M06oh0ZHN+IaXlFbw5ZxOl5RVkZ1rhis9d3cfb54yOjXlwdDcapofOcn35+gGc36OZNyIF4IwOjdj4x9ERZfMkM3n47xIrYely25n77k8sE8vZXZsAMPHy4FnAI2N7eO3nqUmhb3u9Oj75zgxQ4jec3hawZgsb/ziaVY+OBGDast1+/TwDfreyClXN4Riuexstbg+glDD/A0WJB2L6DXf+wJIdGr1hXV944HnPfs0TU1fxh/+tZNifZpGWnEhmWhKX9gu2kd99bieuG9KGpy7vFfKa0xzVHi/p1zIoIsaNUOV+523Mj/hegM5NM0hLTvCGVl7Yu0XIvhn2A+APY3sEHbuoTws2/nE0rbLSSUgQ78Liczfm8+rXvgxcj4krnOI1xjBp1np2hnEu5ztCKUPZqD3OXOfx/FoWgrnrUPBiMLFcjF1RaoKYKvd0h/PRaRa54bS2frVmPFUidx06RnmFCWlCqV8nmScu7cXVg9owto+7Am2Vlc7Ue4Yy/7fnctXAYPNJKCa4OG+z6oY2s3hCGJtkpvLZvT9CREhMEBY9PIKJYR4+GanW+0JVjHQ+jJwPnT9O8xVb8+jZwyGiaMCKGPrzjDWcMfFL11IP05ftYsDjM8ndvJ+8I8W0e3Cay1l813C6B6o6SudEOetPs4LaQjmh3cgvKK5WE5eiVAdRKXcRGSkia0RkvYiMD9HnKhFZKSIrROS9qC7uUE5JDqWVkpTA3AfPdX3PO3O3kBjFwhmPXdITgE/uHhp0rEeL+jSplxbUHo5bhrZj5n0/4r4Rnb1toWLswZcA9ZGddeqhYd0Uv1lKIJ4ErVDO1EgcOVbqrU8fbuTuVMC3vLXA79jKnYf52T+skMYl2w4y+vnQ8eGeCJqNeb5iac769/kFxSFLMcSSyozcBzw+k/Of/SpyR0WpRURU7iKSCEwCRgHdgWtEpHtAn07Ag8CZxpgewC+jubhTRQeOxpMTE4KiQjzkF5a4tjupXyeZzRPH0NNeG7Uq6Ngkw2sKGdG9KTedkROy74vX9mNY52yaRpFx68TzwCuvqLyjcmNeAb0e+cyrXMONTp3HAhXd3+dt8W6XVxjyHElVgSavH7+1gD9OW8WI5772tjlr/wx4fCZ3VONShZEoD/GQrMzIHfDWKFKUk4VoRu6DgfXGmI3GmBLgfeDigD63AZOMMQcAjDF7o7m4iNCnlaV8k1xGs9ef1jaa09Qo+wstpTB7zd6wS+8NzGnIO7cMdv1c4fB0P56B+xRHZUqAw0WhR6c3/m1+yGMpjvIOgTOI9tkZgd29tXk83GrPBDyOzFlRrpH7rwVb+Wpt1aynW1peQWFxGW/O2eRte+JSazaXkpSgNncl7olG87QEnEHB2+02J52BziIyR0TmicjIqC4u8M4tQ3jl+gFeG7WTto3qRnOaGuWuczoCoePiT5QzO1gRMoNyGkboaXHj6b4HYGC993DRMoErOzkdom/P9Y3cnXXlM1OTGNg2i1+db5mmnJFHTjxVLyuz4Mjkhdv5zYfLuOmN0A+dynDr27n0+P0McjdbsfjndWvKdUPasnniGAblZLH/aOTZH/hH2uyPYsaoKLWFqnKoJgGdgOHANcBrItIgsJOI3C4iuSKSC5YJon56MiN7hk6td7OZT7o2dJ2V6iYjNYnNE8eErUZ5IpzRsTEr/nABp0eRVQvw6MU9ef6afgB+VSuvGdyGkrKKqMMhPTb4wP7O6pfL/nABIsJd53Ri88QxYW36OeOn8ov3l0R17fyCYn717x8id6wEX9szgE9XWGGivxvjK76WnZHK5n2FUZ3H+RDMc1noRVFqK9Eo9x2AM6ykld3mZDswxRhTaozZBKzFUvZ+GGNeNcYMNMYMBGvkHgk3m/mY3s2jEPvkJbA2fCR6tqgX1Nbdbuv9yGdh0+0ftx3PD3+8nHkb8+n9yGd+x3ceKiI5UbgoRPRRtDivv3zHIb9EsDHP+y9u4sxCrioaOGaGczfmc+BoKSt3Rq4S6Vzx6/tN0YW+KkptIBrlvgDoJCLtRCQFGAdMCejzX6xROyLSGMtMs5EIHGdAiBJA++wM2jf2mbAu7tuC5gHRQCsCFFnPlvU4u0u21xz2n0U7GOey0Ph/Fu2gtNy4Ooa7Nfd/qAzrnB1Sxv2FJSzbfoiiknIufOFbfj15qXdpwt2H/Z2Vs9dE5bKpFFl1faG1niqak2avj/i+/5vpq280IcwCMrWJguIynv9i3UmbPaxUDRGVuzGmDLgLmAGsAj4wxqwQkUdFZKzdbQaQLyIrgVnAr40xEYc5gT/qUDx8YXdetwt4XR4m/PBUpqGtvDJSk/jLuH50CKh5s2LnIb/9krIKUpISvJUoI+FWefLXF3T22y8uLad78+BZBFhRMxe9+C2r7Vr6n6/cQ4/fz2DmyuBCY876OlXBC7bZysN1Q6y1cqeGWDjFSeAAxLPYeW3mnn8u5tnP13ozqZVTk6hs7saYacaYzsaYDsaYJ+y2CcaYKfa2McbcZ4zpbozpZYx5vyqFvHVoO87r3pTNE8fwzFV9Ir/hFOSgrXw9iWHtGvs7o9/4drPf/to9BSQlJJAZwikayL8dyxRgwmfSAAAgAElEQVR6OLtLE3Ic9W8SRMKuoAXBM4hZjlF612aZQNXXxGnRwH8W86vzu4Ttf98HS7xlje/4UQe/Y9e+5l7u2ElpeUW1mJai5cvV1j1dvuNQhJ5KPKMFNuKEc+y6Ns7oTKfi9qwTC7DO3p66bJdrxMsdP+rAp788y28lKk9SmBMRYeZ9vnr2bjHll/T1t9U/9N/lfvvOhUk8uQ5VUa+mfxufPz8pwf9rnlU3hYzUJM7r1sT1vf9ZtMNb1viVr6yyDs87Rv+7DoVfD+Cp6as560+zaswB++acTeSMn+q9/x5T21vfbdYIn1MYVe5xQuem1qjXuSrT3AfPZaGjBLFnROxcrKSeS6XKuimJdG1Wz69k8LWD27heNykxwatMSsor+PBnp/sd72TLFYrPHWYZj3J6OOABcDw0cBSRc8sI7tIs0y+T1g1n5NBFDif+EwF1+p3kFxR78w1211Di0x+nWfKs3n2YopJyb5VSIGztICW+UeUeJ7jlSmWkJtEoI5XrT7MU84a9BRSXlfP6N1ZizyvX96dRRrCj1FO/5hqHQg9XEvltu15+aXkFA9r64vPvPLsDt9plG6LBGdsfqrbO0u0H+SKKBUGcyVfJicGyZ6QmRUxkctbmERHveeas3+cX7eNkwOMzvbORTfnRhVueKKXl1mcd8/y3dJvwqd+CKXPWH7+PoKy8wm9hGuXkQpV7nBAu4evivlbO2eFjpVz5yly+tX/wnZtmkpggzHvwXD8F7Kl2+eMzc/jk7qGuJhknnkXNBwYslvLrC7qSlpxIh+zoktEeutAXi773SPCot7zCMPbFOdz6di4TPg4/ui91PBzcnMaZae7K3RmyuTfArJKeYpmwDhwtpe+jn4e9PliOzVjgzE14cvpq1z5Tl+4iZ/zUsKajG9+YT9eHP9VVq05SVLnHCf3bZPH4JT2Z/7vggmse08vhojI/u7jHdNGsfprfwiDZ9mheROjZsj43RCgD0aJBHT67dxi/G2OVHGqckep3vnBlGjz8+YrepCYlkm0/KDa5JBnd/4EvKeqduVtC2pP3FRQz135A3Xl2B78FUDxkpiUH1Zc5VlrO2BfnePcve+k7AK+Zwy2LOhLRlD9evuMQax0+kcpQeBzO56MlZbzwpbU+8dLtoRdv/26DdQ8/c4loildyN+/nvGe/qlUzlnV7jhzXA1aVexxx/WltaZIZrMg8C4AcKiqlTUOf0nUqq+TEBL554GyuHdKGX57XOegckejcNNNbn//7357Ll/cP9zv+8Z1nhlWOnpW4/jKuLwA3OGrfbD9wlHOemR0U2vfc52v5t8tyeb/7aJl3+9cXdHW9XnpKIvsKSvwiSt76bjPLXCJMPM+mwFLN0VS73BgmE/ajxdvJGT+VC1/4lvOf+5r/Lg7MDYzMjBW7I/YpDYh37z5hBqt3Ww+TW9/Ojfj+bfuP8p9Flqxu97s28N2GfSGLxFWGK16Zy/q9BbzrKJ4XS75Zl8eI577mn/Mrf99VuZ8CNKqbSnKisGV/obeQ2a8v6BJkR2/dMJ0/XtqLNiGWK4yWxAQJOnef1g1Y/PAI+rVpwOhezfyieppEqJz58ZKdbMwLVpJ/n7eFX09eGtTuqUo5OEx9Hk/m6YUvfMuov3zDnz5dHTJj2hM5c0YH/5WxrnFJ+koKOElZeWiFc++//Esu/PJf0ZVrcOKsA+Tkfkdpak99HUueyic25R0p5vkvrJG+2/2uSvILihnz/Dfe8hHRsHDLAa597XvGhClNXVlqywpjnkHObz9a5ldWOxpUuZ8CpCQl0CE7g/V7CigqKaN783rceXbHGpcjIUH46Odn8tJ1AxjSzqd437/9NO92n1ZBJYmiTrTy8P2m/fb1Qvf52XBf/PqqXYd5afYGvwVPomHR1mCTRnZmKlcOaMXfbrKS7ipb5XLhlsotOv7DtmAZ/nRFb+4+t5M3HNQ5w4g2NNJZMO6vX2/0WySmOmP41+4pYMXOwzz92Zqo3+P5fJ7ZyPHiXNcg3EI3seLdeVsr1V+V+ylCTqO6zNmwj637j5KaHPt/+29H+5ynHkcl4Bdb/+cZq5m+bFdQaOTax0f57YcK9+vTOvhB4eG2Ye1DHnvzx4NCHrusn39B1KMlZVz/+vdMtpO8PCuFdbUzdT1x8tGydPtBNu0rZODjM8kZPzVif0/G9qKHRzCkXUP+e+eZ3hXGPI7wm9/0Ka3AaqAN0t0fnIu3+T9knDOnzfmF7D18jMc/WVnlJQ48jnTnmgCR7N9u4bwAnyzdybOVeEh4kr/At/pbbcItSzwcsf+VKzVC/TrJHCutYO2eAha7jDhrmt6tGpD70Hk8c2UfmtX3+QkSEsRrTpk0a4N3RSgngYtbD396tus1wmWi1ktLJpSft3VWHS7r15L22XX5+fAOfrkCz17dl28eONu7vzGvkG/X7/NWtSyvMCQkCC0b1CE5UVzDMCH0mrQHCks4++nZ7LMdsRUR7MjHyspp17guDeum8K+fnk5fxwMty2XB+MAQ04NHSzlaEmyCSE1KDGrzMHnhdh79ZCWvf7uJ2VHW6o8WTyVRjyL72bsL6frwp2ETx5y2dufD5q73FvP8l5HrB50sfLhoe6Ucq6rcTxHO6tw4cqcapnFGKpe7lE52lud10j67Lh8HLFsIlsL6Zp1PyeQ0SmdsnxZhlzMEuDeE47hjk0yevbovX94/nAdGdg3KBWiVVccb3rnPEQ2Td6SY/MISr+2+tNxQWm7IdVmj1jkyBavERssGdVgcYGYpcFG8Tg4XlYY0W7kp9z12PaefDG3H3fbaBPkFwaaaJ6dbiVHN6wc76D9espMt+ZZpprKjyWg5VFTKVX+dy/TllsM4XB2gcofCKywOHuVHE/niNgMpquRawJv3FbJ4a+XMapXlvfnRm2ZUuZ8i7HVkrkZyYMYaN3PK8C7ZfHn/cO+x5wOKgd3wt/lc97rl4CwtN961aMNx19kdmffgufzyvKDq1GEREa9ZaY+j+N1d71mzjC9W+Ve1DDSFADwekOWalpxA9xb1vM5bD4eOhlee36zb52p3B6iTkkj77Lp+xdxuetNy0A3MyaKfbZPf5xKuOWe9FQbpfECc1ck3QPBEFVV1HaAsh5lo/ibfQ3H7gdAjd+fs5trXg53c0fgZil2S5nYcrJxvYfjTs7nUDp+tKtKSE/jJUN9yo7/7KPrsbVXupwiXOGzFkZKSagNf3O+f1RoYRjnWpb78nPX5FBSXseNgEf9ZFDmsMCFBaFY/jcwQNttweEbLznIPHkeup0LnYxf3AHBdXerDRZaN3rMmb6KIXykGD9GsZhUuxLRni/rsLyzxOh09g9ySckOjutZDPnDk7jTdOPMN+rZuEFSPZ/ryXVW6ALozVBd8JrhwNnCnWSawMB3AGRO/ZHqE++gZ3Z/RoRF/GGv937YfKOLQ0dKQ2dKhON56Prmb9/v5CErKKjhWWnHc51PlforQsG4KSx85n/tGdObcru4Fs2oTHbIz+O1oX4x64FquAEsfOT+orbLRJuC/Zmy0o3hPUbaXZwc7TH91gWXrH9nTV4/Gza4NcN0QK0Fs9+FjNHNJtpoYIsPUQ720JC7tF7jqpY/szFR2Hz5Gl4c+9WsvKaugUYb1EAocuTsTo87r3tS7/fXavKBqo/M27qfLQ5+yfu+JRap4CJwJRKNYr3vdV6nTGYXl5M0IDtJj9nUu7tuCC3pYK8Nt23+UPo9+RueHpru+5y8z13H6k18Etf/vh+MrtXzFK3N5/sv1XhORZy2B/yze4Woei4Qq91OIemnJ3HNup0ov2h0rOjgW4+7ZInhFrnppyd66OWApXM8o+eK+0a8c1bCuz0z14zPahenpI8W+h27FxzwZrdkO89c9//TFsDuTipyVHGf8cljQuUIpK+f7w9X9yXCs6uUc4fZqWZ/Gti8hv7CEwuIycsZPJWf8VJ761Hqg/Ony3jx+SU9v5dBereqHDDe86IU5ru2VpbC43HXNBs8M58yJXzLy/772O+Z58KcmJXiXiwwkJ0LuRrH9f0xLTqRJppUXsnKX77NucakT9NzMtew6dCyoztDxJlN5/o+ekfpL9sDhptPb8t34c7jlzOi+mx5i+iuPtuaIcmpybjffqPGnIUIXH7moh3fbWSsmsA57OEb1bMaQdg156vJe1A8RGhhIuFo+bo7cmav2cNlLc/jrVxv8atb8ZmRXHrmoOw+O6kb99GRm/Wo4jTN8du722RlB5wLrAfHk9FUUlpSHNYskOEKCnIvjdGmWSVpyIhmpSewrKPYzzby/wMqGTE9NpH6dZBZPOJ8HRnbhNyO7cuPpOa7XiVRhMxqW7zjE7sPH2HmwiFdvGOB37FhpBXsPH2PHwSLvA6a8wlBSVuF9mF7Up4XfLKShY/Wt3YdDl4GY8PFyznnmK8B6QCQkCK2z0vmnw3kZOKNwRq1MW+afJXwgysXXnRwt8ZUGyQuYSd13fhdEhDvPjv47DTFW7v+87bTInRQF/JJonCQlJrB54pggc4pn4Y9oz/2vn57O1YPcyxq7EW607KSLo+Txoq0HeXL6ar/ojTopidx8Zjvv52vXuC5dm/kcoG7OToAPF27nr19ZK1mGS25pUs83ezhz4pdBxxtlpPDJ0l1MWx5sk65r5x8kJgg/H96RzLRkRnRvyqYnR3P1wNZB/Y+n/snCLftZYjuEL3zBWkt37sZ8znM82D0M/qPPBLJm9xE6/HYanR+azqGiUtpn16VRRgr5BSVeOZwPtn1hCqS948jy9fhfmgcs8BIYPrzKMapPTPAfrS/bcYhnP19bqRH8Esf5n/t8LS/YGcHg86k0ykjlNXtFumiISrmLyEgRWSMi60VkvMvxm0UkT0SW2K+fRHPeymYeKqce/3d1X351fuRaN/3a+FekjKZY2YniucRTl/fy1sQJ5J+3Bw9giksts8wr1w8IOgZw7RDfQ2Zrvn/ERt4Rq178+P/46ueEW7x8TO/mQbb8B0f5fBlb8o+Sd6TY1bbvWdUrEBHhqSt6B7X3e+xzv5W1PBhjePHLda6ZrZe/PJdLJgWbdJwPc7cyEp68ArCU6ca8QhrUSaGkvIJej3zGsdJyyisquOn0towb1DpoNBwKz7oIaQFx/oGLzDjJO1LMAke46+w1eTz/xTq+3xj9gurO9QdmrtrLM59b9vZeLf3NkSO6Bz/0QhFxjTURSQQmASOA7cACEZlijFkZ0PVfxpi7or6yokTBJWGchU6Gdar5OP7kxATLLNApm5YN6nCoqDQodLFh3RQG5WSxwFHfxWPCCJUp3NGx/u2aPUfo+fsZPH9NX7Iz0rjoxW+D+nuiO9yol5bM3AfPod2D07xtV7mMut1wZg67MbJHMz51FC47eLSUH7+5gKn3DKWHw0eybX8RT3+2lk9X7OaTu8/ytjtDGJ2lh39tO6RvPiOHpAThoQu7B2XrBlb0BOjewprxFBSX0fVhy4GckCBkZ6aSX1Ds6p+Yu8FfAXv8JGsiVOmscMxSnv5srWufp2as4eOO/t/LA4UllBvj9Xe4nc/J5f2j+/67Ec3IfTCw3hiz0RhTArwPXHzcV1SUasA5UneOfKuTPq0sBeaZNt94eg53nxscbfP0lf7r/ubZKfapSe4/v05NMrjn3E783K5/U1Bcxi1v5bJxn3vhqEjr4AbOYtIcJR5+1Dk75PuSk8LPfiZd15+Vj14QpDCnBESLeOzVzjDFwuIyxrzge1Ct23uErPRkOjXJ8NY9emRsDx66sLvrtTfnB88C+rrkRySI0CQzlQpjmbICuea14Lh4CB9XD9ElRv2w7SAHAsIY+z32OQMfnxnkhA2l3OuEmD1FQzTKvSXgrDe53W4L5HIRWSoik0UkuqGBolQhp7W3pu+nt29UI9d77caBvH3LYL+IFDcCa5/c8a6V7BRKuYsI943oHJTM5RYqCe4O3HA4r/v2LYP9kpycdAjhzPWQmCCkpyTRNCApzmN28rBmj6XUjfHZ/Xv8fgarHIupX/va94gIQ9qHjw4Kxcz7hrnG+89Zv89rR3/gQ/+KlhvCVFl0W1/XGe4ZrQO532Of84FLmeQ352z2bpeVV3jNTIEmqDoRZk/hqCqH6v+AHGNMb+Bz4G23TiJyu4jkikiutV9FV1cUfEq0pr5XDdJTwo58PWTVDS4DEA0NA94XmHXbp3UDHg4xsg1HoHM6cGYBVuhgtA+NO4b7R3EEjkKdyVk7DhYF1Zf3cOBoCYmV+Oc5lfmxUvdzHi4q9cbzB3Kpw9b/0Jhu/NdR2uLFa/tz/4jOrH5spLftvGd9IZjO9QY8ZKUn80JA5jTAA5OXesNMPXiK3X25eg/3fvADa/dYD5ofdfH/PjkL6VWWaP57OwDnSLyV3ebFGJNvjPEYzV4HXD1FxphXjTEDjTEDAQTV7krV4XHQH3WpLxJr1j8xKmgt2V4tQ1etBGvZwkvseP3kRMEZfHFmx0Z8fOeZ3Do0utjnwEqaTpzhnysfvYDlf7iAT11i7kNx4+k5fk7vQJPGwLb+o1FnyQYnxoSOivJwXremdG5qzSicC7h7Zhkj7QQkDylJCd6ZXOMAJe+s2X7r0HZ+Zp205ETuPrcTacmJ3v+BT07fP+K2s3z3f/yoriFnY4Fs2X+U5TsOcctbuX5JT56SEB5CObWjIRpJFgCdRKSdiKQA44Apzg4i0tyxOxYIvTy8olQTnpF7dRWzOhGSEhNo3bCOX1tgdctARIT/G9ePi/q0oLTc8J9FPptxYDRHJFKSrJDRzRPHBB1rVi+N0b2a8cbNA0lPSSIjNcnPLh8Nd53j8zV8sy7Pb3nBwJF8uEW7Q5UZePWGAYzq2YzXbxroXRPYmS/gsU2/csMA3vzxIFo2qMOvL+jC6zcNIikxgXGDWoNjMLnX8YA5vX2jsNFV/zfOfzTu/H5dMcA37k1MSPBGbb18Xf+Q5wPYfeiYN/TTSaBJq7L/BycRDTrGmDIRuQuYASQCbxhjVojIo0CuMWYKcI+IjAXKgP3AzdFcXM0ySlXimabXRuUO4cvohsMTUudcai2awmjRkpggvHSde1hmZfjHT4bw8uwNfLt+H8Ofns2fr+hDalJCkHLf4nCGXti7OWnJid56+KFC5c/v0Yzz7VG5Z43fvCPF3HxGDmd08PexnN2lCXPGn+PX1qJBHfYVFHOstJy05EROd8T8bzsQuUBYk8xU9h4pZmv+Ua/eGtG9KV2aZdKmYTpb9x8lLTmB7MxU7wP0jh91CFnPf2uIBU9O79CI/955Jr+ZvJT8wmI6RvB7hCOqOYQxZpoxprMxpoMx5gm7bYKt2DHGPGiM6WGM6WOMOdsYE9WSNqrblapkuG2vPF6nXE3wv7uGVvo9A9pmBbUtCVEJMpac2bGxV9Yjx8q4492F/PitBSzfYTlOP7nb+uyetPrB7RryxKW9/Gz+Pz4zJ+J1DL4nwCNje3iVfjhaNrBmTe/O22KFIzpsXJEcx+BzWt/85nxvJrSnpo/HFBP48PaUVA7ELaoHrPuTlpxI39YNmHHvMHIfGhF1xrQbJ0eREUWJgj6tG7D+iVGc1SmykzNWtG1c+fVpH704uIpnXpiMy1hSNzV4duIJjezU1F+JZmekBkW4jHDJTA1kTO8WjOndnHkPnhu1XC1t+/zjU1f5JRwlJkhUWZ8X9rEszxv3FTLaXqu1rh0ldaYdyx5Y3KtuahKbJ45h/ROjuMwRrx5q9aueLYPrJ50IMVXuNZFFqJxa1PaiaBl2aJvbaDwU2ZmpfH6vv4PzOGtTVTtui2V4CIyEOeJSCz45CodkRmoSk67t77eCVySaOsJInTb/5vXTIvo+AO4YFlzXJcN+kP12tBVpE0o5JyUm8Iw9O+nevJ6rv8RZpqKqqN2/BEWJMxIShE/uHhp2nVY3PJmTnoW9nVmstYm7QpgiwBolOys+jh/pK4Ow4Y+jeffWIQxyKTVQFbR11Il/21FLJtrFRtzCWTNSrRF4SlJCSFOLBxFh88QxTL1naNDqYz8d1p4Z90YfnRQtxx8hXwXouF05FTme6XeD9BQWPTyC+nWSGd45m3a1tKJqcmICd57dgUmzgh2JIsLlA1p6FyrxlAsAS/EPrcYSEqFCLJ2VRCtL++P4H4gII7o35bGLe/Dwxyt455bBDIsiV+J4iLFZJpZXV5STi4Z1U0hMEIa0b0STzMov3lBTeMoCN0hP9hZU62TPNDwZmBOOI/nqRAk0bUHlaq97Sgv/ZVxfNj05utKZwU6uP60tH/38DL+lC6saNcsoilKlNK2Xxid3D+Xb35zjzbL12LU9JZpviTL5qirp1DTTL+MU4K83RB8Ces1gq2bRoJyGJ+wvFBH6tcmqVr9jbM0yOnRXlLjEY3rypM/XFke3MynoqoGtvEvqRcOY3s0Z0zs4Cay2ErM73jbCsleKopz8eBy/Px9euVWEaoLhXWr/WsInQsxG7oGV8hRFiT8apKe4ljyoDYzu1Txyp5OYmJplFEVRappbh7ajLERlynhClbuiKKcUx1Mm+WSkdng5FEVRlCpFlbuiKEocospdURQlDlHlriiKEoeoclcURYlDVLkriqLEIVEpdxEZKSJrRGS9iIwP0+9yETEiErn6vaIoilJtRFTuIpIITAJGAd2Ba0QkKFBURDKBXwDfV7WQiqIoSuWIZuQ+GFhvjNlojCkB3gcudun3GPAUcMzlmKIoilKDRKPcWwLbHPvb7TYvItIfaG2MmRruRCJyu4jkikhuXl5epYVVFEVRouOEHaoikgA8C9wfqa8x5lVjzEBjzMDs7Nq7iLGiKMrJTjTKfQfQ2rHfym7zkAn0BGaLyGbgNGCKOlUVRVFiRzTKfQHQSUTaiUgKMA6Y4jlojDlkjGlsjMkxxuQA84CxxpjcapFYURRFiUhE5W6MKQPuAmYAq4APjDErRORRERlb3QIqiqIolSeqkr/GmGnAtIC2CSH6Dj9xsRRFUZQTQTNUFUVR4hBV7oqiKHGIKndFUZQ4RJW7oihKHKLKXVEUJQ5R5a4oihKHqHJXFEWJQ1S5K4qixCGq3BVFUeIQVe6KoihxiCp3RVGUOESVu6IoShyiyl1RFCUOUeWuKIoSh6hyVxRFiUNUuSuKosQhUSl3ERkpImtEZL2IjHc5foeILBORJSLyrYh0r3pRFUVRlGiJqNxFJBGYBIwCugPXuCjv94wxvYwxfYE/Ac9WuaSKoihK1EQzch8MrDfGbDTGlADvAxc7OxhjDjt26wKm6kRUFEVRKks0a6i2BLY59rcDQwI7icidwH1ACnBOlUinKIqiHBdV5lA1xkwyxnQAfgM85NZHRG4XkVwRyc3Ly6uqSyuKoigBRKPcdwCtHfut7LZQvA9c4nbAGPOqMWagMWZgdnZ29FIqiqIolSIa5b4A6CQi7UQkBRgHTHF2EJFOjt0xwLqqE1FRFEWpLBFt7saYMhG5C5gBJAJvGGNWiMijQK4xZgpwl4icB5QCB4CbqlNoRVEUJTzROFQxxkwDpgW0TXBs/6KK5VIURVFOAM1QVRRFiUNUuSuKosQhqtwVRVHiEFXuiqIocYgqd0VRlDhElbuiKEocospdURQlDlHlriiKEoeoclcURYlDVLkriqLEIarcFUVR4hBV7oqiKHGIKndFUZQ4RJW7oihKHKLKXVEUJQ5R5a4oihKHqHJXFEWJQ6JS7iIyUkTWiMh6ERnvcvw+EVkpIktF5AsRaVv1oiqKoijRElG5i0giMAkYBXQHrhGR7gHdFgMDjTG9gcnAn6paUEVRFCV6ohm5DwbWG2M2GmNKgPeBi50djDGzjDFH7d15QKuqFVNRFEWpDNEo95bANsf+drstFLcC090OiMjtIpIrIrl5eXnRS6koiqJUiip1qIrI9cBA4M9ux40xrxpjBhpjBmZnZ1flpRVFURQHSVH02QG0duy3stv8EJHzgN8BPzLGFFeNeIqiKMrxEM3IfQHQSUTaiUgKMA6Y4uwgIv2AvwJjjTF7q15MRVEUpTJEVO7GmDLgLmAGsAr4wBizQkQeFZGxdrc/AxnAv0VkiYhMCXE6RVEUpQaIxiyDMWYaMC2gbYJj+7wqlktRFEU5ATRDVVEUJQ5R5a4oihKHqHJXFEWJQ1S5K4qixCGq3BVFUeIQVe6KoihxiCp3RVGUOESVu6IoShyiyl1RFCUOUeWuKIoSh6hyVxRFiUNUuSuKosQhqtwVRVHiEFXuiqIocYgqd0VRlDhElbuiKEocEpVyF5GRIrJGRNaLyHiX48NEZJGIlInIFVUvpqIoilIZIip3EUkEJgGjgO7ANSLSPaDbVuBm4L2qFlBRFEWpPNEsszcYWG+M2QggIu8DFwMrPR2MMZvtYxXVIKOiKIpSSaIxy7QEtjn2t9ttiqIoSi2lRh2qInK7iOSKSG5eXl5NXlpRFOWUIhrlvgNo7dhvZbdVGmPMq8aYgcaYgdnZ2cdzCkVRFCUKolHuC4BOItJORFKAccCU6hVLURRFOREiKndjTBlwFzADWAV8YIxZISKPishYABEZJCLbgSuBv4rIiuoUWlEURQlPNNEyGGOmAdMC2iY4thdgmWsURVGUWoBmqCqKosQhqtwVRVHiEFXuiqIocYgqd0VRlDhElbuiKEocospdURQlDlHlriiKEoeoclcURYlDVLkriqLEIarcFUVR4hBV7oqiKHGIKndFUZQ4RJW7oihKHKLKXVEUJQ5R5a4oihKHqHJXFEWJQ1S5K4qixCFRKXcRGSkia0RkvYiMdzmeKiL/so9/LyI5VS2ooiiKEj0RlbuIJAKTgFFAd+AaEeke0O1W4IAxpiPwHPBUVQuqKIqiRE80I/fBwHpjzEZjTAnwPnBxQJ+Lgbft7cnAuSIiVSemoiiKUhmiUe4tgW2O/e12m2sfY0wZcAhoVBUCKoqiKJUnqSYvJiK3A7fbu8Uisrwmr3+CNAb2xVqISnAyyXsyyQoqb3VzMskbC0k8xTwAAAUaSURBVFnbRtMpGuW+A2jt2G9lt7n12S4iSUB9ID/wRMaYV4FXAUQk1xgzMBohawMqb/VxMskKKm91czLJW5tljcYsswDoJCLtRCQFGAdMCegzBbjJ3r4C+NIYY6pOTEVRFKUyRBy5G2PKROQuYAaQCLxhjFkhIo8CucaYKcDfgL+LyHpgP9YDQFEURYkRUdncjTHTgGkBbRMc28eAKyt57Vcr2T/WqLzVx8kkK6i81c3JJG+tlVXUeqIoihJ/aPkBRVGUOCQmyj1SOYNYICKbRWSZiCwRkVy7raGIfC4i6+y/WXa7iMjztvxLRaR/Dcj3hojsdYaPHo98InKT3X+diNzkdq1qlPcREdlh3+MlIjLacexBW941InKBo73avysi0lpEZonIShFZISK/sNtr5f0NI29tvb9pIjJfRH6w5f2D3d7OLleyXqzyJSl2e8hyJqE+Rw3J+5aIbHLc3752e8x/b64YY2r0heWU3QC0B1KAH4DuNS2Hi1ybgcYBbX8Cxtvb44Gn7O3RwHRAgNOA72tAvmFAf2D58coHNAQ22n+z7O2sGpT3EeBXLn2729+DVKCd/f1IrKnvCtAc6G9vZwJrbZlq5f0NI29tvb8CZNjbycD39n37ABhnt78C/Mze/jnwir09DvhXuM9Rg/K+BVzh0j/mvze3VyxG7tGUM6gtOMsqvA1c4mh/x1jMAxqISPPqFMQY8zVWJNKJyHcB8LkxZr8x5gDwOTCyBuUNxcXA+8aYYmPMJmA91vekRr4rxphdxphF9vYRYBVW1nWtvL9h5A1FrO+vMcYU2LvJ9ssA52CVK4Hg++tWziTU56gpeUMR89+bG7FQ7tGUM4gFBvhMRBaKlUkL0NQYs8ve3g00tbdry2eorHy1Qe677KnrGx4zRxi5alxe2wTQD2u0Vuvvb4C8UEvvr4gkisgSYC+WktsAHDRWuZLAa4cqZxIzeY0xnvv7hH1/nxOR1EB5A+SK6e9NHao+hhpj+mNVv7xTRIY5DxprnlVrQ4tqu3w2LwMdgL7ALuCZ2Irjj4hkAB8CvzTGHHYeq43310XeWnt/jTHlxpi+WBnug4GuMRYpLIHyikhP4EEsuQdhmVp+E0MRIxIL5R5NOYMaxxizw/67F/gI6wu4x2Nusf/utbvXls9QWfliKrcxZo/9o6kAXsM3pY65vCKSjKUo/2GM+Y/dXGvvr5u8tfn+ejDGHARmAadjmS88uTbOa3vlEv9yJrGUd6RtDjPGmGLgTWrh/XUSC+UeTTmDGkVE6opIpmcbOB9Yjn9ZhZuAj+3tKcCNtpf8NOCQY/pek1RWvhnA+SKSZU/Zz7fbaoQAv8SlWPfYI+84O0qiHdAJmE8NfVdse+7fgFXGmGcdh2rl/Q0lby2+v9ki0sDergOMwPITzMIqVwLB99etnEmoz1ET8q52POgFyz/gvL+17vdWI17bwBeWd3ktlt3td7GQIUCe9lhe+B+AFR6ZsOx8XwDrgJlAQ+Pzpk+y5V8GDKwBGf+JNdUuxbLd3Xo88gG3YDmi1gM/rmF5/27LsxTrB9Hc0f93trxrgFE1+V0BhmKZXJYCS+zX6Np6f8PIW1vvb29gsS3XcmCC43c3375X/wZS7fY0e3+9fbx9pM9RQ/J+ad/f5cC7+CJqYv57c3tphqqiKEocog5VRVGUOESVu6IoShyiyl1RFCUOUeWuKIoSh6hyVxRFiUNUuSuKosQhqtwVRVHiEFXuiqIoccj/A7/JQvFannB/AAAAAElFTkSuQmCC\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.6985253691673279.\n", "Better model found at epoch 2 with accuracy value: 0.7099945545196533.\n", "Better model found at epoch 3 with accuracy value: 0.7165483236312866.\n", "Epoch 7: reducing lr to 1.4996633150096014e-06\n", "Better model found at epoch 8 with accuracy value: 0.7192791104316711.\n", "Better model found at epoch 10 with accuracy value: 0.7252867221832275.\n", "Better model found at epoch 11 with accuracy value: 0.7362097501754761.\n" ] } ], "source": [ "learn.fit_one_cycle(epochs_all, max_lr = slice(1e-6, 1e-5), wd=0.05, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_stage2_resnet50_fold1'),\n", " ReduceLROnPlateauCallback(learn, monitor='accuracy', \n", " mode='max', patience = patience)])" ] }, { "cell_type": "code", "execution_count": 37, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "1831\n" ] }, { "data": { "text/plain": [ "True" ] }, "execution_count": 37, "metadata": {}, "output_type": "execute_result" } ], "source": [ "learn.load('Best_model_sz224_stage2_resnet50_fold1')\n", "\n", "interp1 = ClassificationInterpretation.from_learner(learn)\n", "print(interp1.confusion_matrix().sum())\n", "\n", "losses,idxs = interp1.top_losses()\n", "len(data.valid_ds)==len(losses)==len(idxs)\n" ] }, { "cell_type": "code", "execution_count": 38, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "interp1.plot_confusion_matrix()" ] }, { "cell_type": "code", "execution_count": 39, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.6537219451416985" ] }, "execution_count": 39, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix1 = interp1.confusion_matrix()\n", "\n", "FP = confusion_matrix1.sum(axis=0) - np.diag(confusion_matrix1) \n", "FN = confusion_matrix1.sum(axis=1) - np.diag(confusion_matrix1)\n", "TP = np.diag(confusion_matrix1)\n", "TN = confusion_matrix1.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "TPR.mean()" ] }, { "cell_type": "code", "execution_count": 40, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 00:44

" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "0.6699866694639852" ] }, "execution_count": 40, "metadata": {}, "output_type": "execute_result" } ], "source": [ "interp1_tta = ClassificationInterpretation.from_learner(learn, tta=True)\n", "confusion_matrix1_tta = interp1_tta.confusion_matrix()\n", "\n", "FP = confusion_matrix1_tta.sum(axis=0) - np.diag(confusion_matrix1_tta) \n", "FN = confusion_matrix1_tta.sum(axis=1) - np.diag(confusion_matrix1_tta)\n", "TP = np.diag(confusion_matrix1_tta)\n", "TN = confusion_matrix1_tta.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "TPR.mean()\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold 2" ] }, { "cell_type": "code", "execution_count": 41, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "this Learner object self-destroyed - it still exists, but no longer usable\n" ] } ], "source": [ "gc.collect()\n", "learn.destroy()\n", "seed_everything(SEED)\n", "bs = 32\n", "data = get_data(metadata_train_valid2, bs, 224)\n", "\n", "#Define Convolutional Neural Network with backbone resnet50\n", "learn = cnn_learner(data, models.resnet50, metrics = [error_rate, accuracy],\n", " callback_fns=[ShowGraph])\n", "\n", "learn.loss_func=nn.CrossEntropyLoss(weight=class_weights)\n", "\n" ] }, { "cell_type": "code", "execution_count": 42, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Learn rate finder\n", "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 43, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 02:53

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
01.4677291.0157990.3389260.66107400:43
11.0244680.5976180.1968680.80313200:42
20.8338930.5759280.1711410.82885900:43
30.6711420.5589270.1784120.82158800:42
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.6610738039016724.\n", "Better model found at epoch 1 with accuracy value: 0.8031319975852966.\n", "Better model found at epoch 2 with accuracy value: 0.8288590312004089.\n" ] } ], "source": [ "#Fit heads\n", "learn.clip_grad(1.0);\n", "lr = 1e-3\n", "learn.fit_one_cycle(epochs_head, max_lr = lr, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_resnet50_fold2')])\n" ] }, { "cell_type": "code", "execution_count": 44, "metadata": {}, "outputs": [], "source": [ "learn.load('Best_model_sz224_resnet50_fold2')\n", "learn.unfreeze()" ] }, { "cell_type": "code", "execution_count": 45, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 46, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 14:33

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
00.6817180.5552750.1661070.83389300:57
10.6280090.5462180.1672260.83277400:58
20.6302180.5827550.1767340.82326600:57
30.5849940.5882260.1851230.81487700:58
40.5421880.5572750.1873600.81264000:57
50.4504360.5294610.1672260.83277400:57
60.4047080.5605230.1661070.83389300:57
70.3566810.5905610.1840040.81599600:58
80.2974160.6007750.1638700.83613000:57
90.2853700.5810900.1571590.84284100:57
100.2449330.5919170.1549220.84507800:57
110.2317760.5664620.1521250.84787500:57
120.1992750.5612600.1465320.85346800:57
130.1927460.5791870.1510070.84899300:57
140.1712760.5749820.1538030.84619700:57
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvqOYd8AAAIABJREFUeJzt3Xd8ldX9wPHPyR5kkISdQMImbAhDGYI4GBW0KuJelWq1altHHHVVLdZqW/2hlroH4qgVFBCRIaCABBmyCRAg7EAWITvn98dz98i9SW7Wzff9et0Xzzj3uSeX5Hufe8b3KK01Qggh/EtAY1dACCGE70lwF0IIPyTBXQgh/JAEdyGE8EMS3IUQwg9JcBdCCD8kwV0IIfyQBHchhPBDEtyFEMIPBTXWCyckJOjk5OQaPeeXI/kEBwbQu31U/VRKCCGauI0bN+Zordt4KtdowT05OZmMjIyaPSd9IQAZs6bUR5WEEKLJU0od9KZcozXL7D1xlk83HG6slxdCCL/WaMG9pKKSh/67tbFeXggh/Jp0qAohhB9qtDZ3IYSoqfLycrKzsykpKWnsqtS7sLAwEhMTCQ4OrtXzm01wr6qSvPNCtHTZ2dlERUWRnJyMUqqxq1NvtNacPn2a7OxsUlJSanWNZtMsU2WzqEh5ZZXLMlprktMX0vOxxQ1VLSFEAyopKSE+Pt6vAzuAUor4+Pg6fUNpRsHdur0mM8dlmQpToTI3wV8I0fz5e2A3q+vP2YyCuzW6BwW4/qFty1RIgBdCtGDNJrgfOnPOsl1R6dz+XlJeydNf7bDs/3Ikv0HqJYRoOfLy8njttddq/LzJkyeTl5dXDzVyz6vgrpSaqJTarZTKVEqluzj/D6XUZtNjj1LK5z/FJf9YZdk+UWDfDpWVU0TvP3/D3PWHLMeO5ft/b7oQomG5C+4VFRXVPm/RokXExsbWV7Vc8hjclVKBwGxgEpAKXKuUSrUto7X+g9Z6kNZ6EPAq8IUvK6m1/Z36S0v32O2v23/a6Tm/++hnX1ZBCCFIT09n3759DBo0iGHDhjFmzBimTp1KaqoREi+//HKGDh1K3759mTNnjuV5ycnJ5OTkkJWVRZ8+fbjjjjvo27cvl1xyCcXFxfVSV2+GQg4HMrXW+wGUUvOAacAON+WvBZ70TfUMpRX27eenCkvt9s+WVv+pKYTwP09/tZ0dRwt8es3UjtE8eVlft+dnzZrFtm3b2Lx5MytXrmTKlCls27bNMlzx7bffJi4ujuLiYoYNG8aVV15JfHy83TX27t3Lxx9/zH/+8x+mT5/Of//7X2644Qaf/hzgXbNMJ8A2CUy26ZgTpVQXIAVY7ub8TKVUhlKqRhnDCkrKqz2/+XDDtmUJIQTA8OHD7cahv/LKKwwcOJCRI0dy+PBh9u7d6/SclJQUBg0aBMDQoUPJysqql7r5ehLTDOBzrXWlq5Na6znAHIDQDj00wEfrD3L9iC7VXrSwxPnO/Fh+MR1iwgH4eusxp/NDOjds+5YQomFVd4fdUCIjIy3bK1eu5LvvvmPt2rVEREQwbtw4l+PUQ0NDLduBgYH11izjzZ37ESDJZj/RdMyVGcDHNanAY//b5rFM3rkyp2NnipyP2Soqdfn5IoQQtRYVFUVhYaHLc/n5+bRu3ZqIiAh27drFunXrGrh29rwJ7huAHkqpFKVUCEYAX+BYSCnVG2gNrPVlBUsrKrnydeslH5/SBwDtIRvBGRcfCEIIURfx8fGMGjWKfv368eCDD9qdmzhxIhUVFfTp04f09HRGjhzZSLU0eGyW0VpXKKXuAZYAgcDbWuvtSqlngAyttTnQzwDmacehLXW08WCuZXt4Shy9TKswnSsz7syPm4Y8xkYEU1peRXF5JdPTEvk0I5vnFu4gfVIfAt1MehJCiJqaO3euy+OhoaEsXuw69Ym5XT0hIYFt26ytFQ888IDP62fmVZu71noRsMjh2BMO+0/5rlpWCmtg/mTmSH4+ZAT7wpJyvtl2nM83Gn29vdpFERocyKo9pwgNCgTgP6sP8KsBHRmYJO3vQoiWpclnhTxZaNyZK2XkWggPNqr88tI9bLcZBnWurJK3bhnG1uw8Vu2x5p75bucJCe5CiBanyacfuG/eZgBmXzcEgIgQ465878mzduWUglahQZzfLYG2UdbeaFcjbRxtP5pPcvpCMk/ad5ScLa2guEw6ZoUQzU+TD+5m5pZ8c3Avc5jYNMjm7rxnuyjLtjcZIpdsOw7Agi32Qyr7PbmEMX9zOWRfCCGatGYT3IvKjDvwyFD7lqSoMGP/oYm9LcdG90jg5ekDiY0IpsiL2auhwcYHxrnSCioqq/ho/UFLVsmcszLqRgjR/DT5NnezC3u3BZyDu7nZJTTI/nPq10MS+c/qA14F95BA47lvrjnAm2sOAPDpBuuk3PzicmLCq1/qqqS8kpW7TzGxX3uPryeEEPWt2dy5J7QKrfZ8cKDzj7LzWAHf7TzJ/fM22R3PLy7n/bVZloRkIUHOz92SbU0ZPPN9z9kSZi3exZ0fbmTjwTMeywohWoZWrVoBcPToUa666iqXZcaNG0dGRo0ysnilyQb3c2UVHMsvJjBAcc/47nW61pebj5JrM6P1yfnbeGL+dn46YATi0x5mu64/4DlgH8gpAqCg2PM3hZvf/omb3v7JYzkhhH/o2LEjn3/+eYO+ZpMN7n/4ZDPn/XU5lVWa3FrONr12uDVrwldbj1q284uNRGSFJRVorXllmXNyn5oyrwKVV1x9XXcdL+D7PadYtecUhR4Sogkhmpb09HRmz55t2X/qqad49tlnmTBhAkOGDKF///7Mnz/f6XlZWVn069cPgOLiYmbMmEGfPn244oorGjXlb6P4cZ81R/tH6w/x3BX9ncr8a8Yg7pu3mW5tIp3OAbSLDrNsR4RYf1RzE05ZZZXdDNi62H/KuHP/wydbuGJwossye04UMvGfqy37/Z/6licvS+XWUbVb3VyIFm1xOhz/xbfXbN8fJs1ye/qaa67h/vvv5+677wbg008/ZcmSJdx7771ER0eTk5PDyJEjmTp1qts1UF9//XUiIiLYuXMnW7duZciQIb79GUyaxJ37N6ahiNuO5FNuGqVi24HZ3iZIA9x/UQ8GJsVSYOpM3WcKrI6ycqzHSyus49XNo2MKist9ttju+d2MnM1XDnEd2B3rY/bOD1k+eX0hRP0bPHgwJ0+e5OjRo2zZsoXWrVvTvn17Hn30UQYMGMBFF13EkSNHOHHihNtrrFq1ypK/fcCAAQwYMKBe6tok7tzv/HAjX90zmsv+bw3xkSFcPriTJcgDPDPNPrXn/Rf15P6LevLjvhzHS9mxHeO+fOdJS2ph8+iY9C9896lfaWqWMQ/NdGXmBxudjrn71iGE8KCaO+z6dPXVV/P5559z/PhxrrnmGj766CNOnTrFxo0bCQ4OJjk52WWq34bWJO7cAZbuMO7eTxeV8daaA5wosK625G6s+cgU42555tiuLs8/8Svrh8KyXSct2y4G1rjUt2M0/TpFAzDw6W8pKXc/W/WcKcWwN0MvbZ05V0527jnPBYUQTcI111zDvHnz+Pzzz7n66qvJz8+nbdu2BAcHs2LFCg4ePFjt88eOHWtJPrZt2za2bt1aL/VsMsH9leWZbs9Fh7u+Gw4IUOx/fjKPTOrt8nz7mDCyZk2x7JuHPhaXO89ajTbdcc/6dX8SW4eTPqk3C+8dw8EcI/DmF5ez3OYDwtE5U+C3zTNfUl5JztlSd08BYMvhPEa/sKLGHwpCiMbRt29fCgsL6dSpEx06dOD6668nIyOD/v378/7779O7t+t4ZHbXXXdx9uxZ+vTpwxNPPMHQoUPrpZ5NolnGk0v7up8YFOBFOt8ABVXaWHt1+PPLXJa5Y0xXXlq6h/O7JTBjeGfL8T9e0pOnvzKWiw03tdW7UmyaQWs7rPKGN9eTcTDX7gPGnTNFZU4TtIQQTdMvv1ibdBMSEli71vUyFmfPGjmwkpOTLal+w8PDmTdvXr3XscncuVc3ScnVBKWaeGn6QADu/2Sz2zJ3juvGqgfH0zk+wu541zatLNu3vruB5PSF/PlL59WjzCs/2a7nmmEzEse2QxegQ4x9J7GnO3whhKiJJhPcXQW3sT3beHXX60lcpPHBYTu80tZ9E3oQHBjgFNgBIkOc79Y/WOfcplZs0x7vuF5JWUUVJWXVJzDztGygEELURJMJ7q7ER4bU+3UGJsbwh4t7uj3ft2MM43u18fga58qsbeYFDmmGr/vPOsud+28vMDp/yyqq6GozUmb13upH/gghDD5e7K3JquvP2ajBvXf7qGrPPz3NN6ubx7dyDu4Pm7JIfnbn+dU+NzwkkLdvGebxNc6VVZJgeh3z0n9mGQdzLW3xCaZvEWUVVUxPs86gfffHLI+vIURLFxYWxunTp/0+wGutOX36NGFhYZ4Lu9GoPXgzx3alT4doJv1rtdO5q4cmEh1WfSZGb8W5uHO/a1w37hrXzavnu5roVFpRye3vZnDXuG4ktY6gsKSCQT0SWL03hwM5RU7NTOaf0Tw5a1T3BK4b0ZlZi3fV9McRosVKTEwkOzubU6dONXZV6l1YWBiJie4nRXrSqME9QCn6dIh2ec5VuoHaCg0K5HfjuvHayn0ALPvTBXW+5ivL9rImM4c1mdbmlCDTyJ0TBSW8uvywy+eFhwSy4oFxdIgJs0ymMjtXVmGXJkEIYS84OJiUFEnX4Q2vmmWUUhOVUruVUplKqXQ3ZaYrpXYopbYrpVwvD+70HPv9347tygOX9OTr3492mYa3LgYkxli2u9mMgPHWZQM72u3PXrHPqUyV6Zvikwu2M9WhvFlwoCIlIZKw4EACAhSvXjvYcu6Nlc7XFEKI2vB4m6iUCgRmAxcD2cAGpdQCrfUOmzI9gEeAUVrrXKVU25pU4pVrB9MxJoy05Lia1b4GBiTWbZHsV68dTHRYEB+tP+S2jO2CIY5rvJoFBdh/aF02sCO//9jIN//K8kwmD+jAs1/v5M2b0wirZly9EEJUx5vb4+FAptZ6v9a6DJgHTHMocwcwW2udC6C1dj+V04WpAzvWa2AHz4t9eOO5K/rz858vdnv+ictSuWOM8ZXx843ZAKx4YJxdmaBA5/b7Zy/vZ9me+M/VrMnMYdOhPKdywo9VVUHBMTi4FjZ/DCv+CmtnG1kPqzyvAyyEI28aeDsBtg3I2cAIhzI9AZRSPwCBwFNa628cL6SUmgnMBAhp391nGRm9ERIUwNVDE5k6yHVzibdcdc6aJbaOoKjMfrJSXIR9eVcTssxLCPrSC9/son10GDefn+zza4taKiuC3IOQm+X8yDsIFbajrBRgaucLj4Pk0ZAy1ngk9HRu0xTCga9674KAHsA4IBFYpZTqr7W2u/3UWs8B5gCEdujR4GOZXrx6oM+vufqh8Yz52wrL/s3nJTPXpukmOjyIfp2i2XakADCGQDpynK0K2GXFrI3XTe33EtwbUFUVFB5zHbxzs6DI4QttSBTEJUObntDzEmidbHqkQEwSFJ2CrNVwYJXx2LnAeF6rdpA8xhTsxxjlJdgLB94E9yNAks1+oumYrWxgvda6HDiglNqDEew3VHfh5vrr+NbNadz+nrHmYVJcBH06RFsmKfWyGbt//0U9UErx+Z3n0/vPxhcZV7NglVL8blw33vh+n6VT9lShpCNo0rSGY5vhl88hZw+cOQB5h6DS5v9NBUBMohGwe020Cd7JRkAOb119UI7pBANnGA+tjQ+IA6usAX+badm2mCQj0JsDfkynevuxqayAgmw4sx9O7zN+7jP7jUfhMUhMg9TLofevIDK+/uohPPImuG8AeiilUjCC+gzgOocyXwLXAu8opRIwmmn2e7pwc73ZmNCnnd3+4vvGuCx35wXGOHrbjlF3I3UiQ4MsgR3gT59t4cqhNmNcq6qgNB/OnYHiPCjONT3OWLZ18Rk4l4sqzmV5iNHmX7h4A1Hn3Q6xSQgfKM41AvrP7xnt4YGh0KYXtEuF3pPtA3hMEgT6Zq4GSkFcivEYerMR7HP2WO/qdy+CzR8ZZeO6GXf05oDfqobNfpXlxgeVOWjbPnIPQpXN8pBB4RDXFRJ6QJfzYf9K+Ope+PoPRh1SL4c+l0Fkgm/eB+E15c1ML6XUZOCfGO3pb2utn1NKPQNkaK0XKKPx/CVgIlAJPKe1rjbtWWiHHvqLJauYMqBDnX+IxvDk/G3sPFbIp3ee53Tui5+zeWXZXpb/aZwla2VVlUYpFxOicjLh6M/8tCOTH7ftJYYiYtVZYjnL+M5BNoE7D0U1/1ehMRwqDiEwMo5OHTuxYE8xsZxlTOA24zV7XAJDb4UeF0OA8WFzoqCEJduPc+PILg3a/9HsaA0Hf4Cf34cd84228Q4DYchN0O8qCK/bSCyfqKqCk9utwf7gj1BqNAXSpo+1CafLKIiIg4pSI1Cf2Q9n9tkH8LzDoG36jkJamT5YuhmB3PYR1d7+Lk1rOL7VeJ+2f2lcWwUafQZ9L4fel0Erz+k8hHtKqY1a6zSP5RprGm9zD+51dnQTrH4Zdn4FNkG7QEeQpyPJJYoB3ZNREa0hvDW78oP4dHsR8W3acffk4Xyy7Sz/3nCG564dQ6cOHfl6+0n+9s1uwEi4tmqPMYNv/vVJDDw5HzZ9AGdPQHSiEZSG3Ejy88YQzO/+OJbubaP4ZMMhdh0v5M9TUjmce44u8S18lajCE7BlLvz8gRGkQmNgwNUw+EboOKixa1e9ygo4tgWyzMF+LVQUA8oIyIXHsf29IzQG4rs6B++4rhDZpnZfs7WGE9uMIL/jSzidaTRVdRllBPo+U2v+raI5KM41vtUFBENolM0jGgLr3s3ZLIL7/75dxeT+LSi4m+8AV78E+5Ybf1DDfwP9p/PjccWNc3dTibUJZ+UD40hOMALs0L8steSnyZo1hd+8l8F3O92v02j25k1pXJTazviqvXsxbHzHeG0VyJKKwcytnMDv75hJWkoCyekLAejTIZqdxwpY9qcL6NamFYUl5azem0NQgOKSanLr+4WqSshcZjS77PkGqiqg8/nGB2LqNAhx7jNpFirK4MhGI9DnHoDYLvYBPCKufttJtYYT240gv/1LOL3XGuhTpxmBPqqd5+s0NVobTViH1sGhtXB4PZzc4b58cIRDwDcFfadj7o5Ho1q18Sq4N+pc9xbTEKA17P3WCOqH10NEAkx4EobdDmHGzNnYigIqsV+NKr/Y2rZpDuyXpBp/AG2ivBu3b7lGYDCkTjUeZ/bDxvcY9sPbXBqYQf6893lTT6AN53OKWHYeM77OHz5zjm5tWvHbDzZa0iX7IgVzk5R7EDZ9aLRbFxwx/o9G/s4I6gk9Grt2dRcUAl3OMx6NQSlo3894jH8MTu60BvpFD8CiB402+9TLjd/RqCZ6E1FVaXwbMQfzQ+uh8KhxLjQakoZD319DpyHGz1xa6PAocD6Wm2U9XlJg3yRWB40b3P09uldVwvb/wZp/GL8QMUkw6UUYfIPTHWBClHU8/MS+7flm+3GKTGmEbYdFtjItB5jgItOlKwUl5c4H47rCxU8zfdMoeuWt5rrKZfwm8ENuDv2YpVVDmVs5gR+q+lJWUcXJghK3efCbvYpS2LXQaEvfv9I41v0imDgLek40AqLwPaWMDuh2qTD+USPQm5tuFj8Iix+CzudZm26iG/HbfVkRZGcYwfzwOji8AcoKjXPRnYwPpM4jjUfbVEt/Vq1pbfTpuPswKCmAp3/r1aUkS1V9qCiFLfPgh38ad8nxPeDy16H/1W5HT9hOdrp7fHe+2X6cD9cd5PxuCZbFtwHyzxnB2jGNgTsFxe7XZs08U04mI1lYNZKUimNcG7icqwK/Z3LgT2RVtaN4541sK77C7jnH80to72JcfrNycpfRB7HlYzh32vjQHZcOg66XUUWNoW0f4zH+EeP/Zsd8U6B/CBY/DEkjjCGW4bEQFmsMIQ2PhTDTv+GtjW/AdQ2sYPSzHF5nujNfZ/Rb6EpAQbu+MGC68cHTeWT9/K4oBcHhxsNtf0SzCO5+duteVgQb34MfXzW+qnUYCNPfN8b8evjFC7KZuRoRapRd9MtxSisqeXy+dVm/3HNG80xZpeevbq1Cg+yadmydLLTPOX9Ad+D5iut5qeJqLg3YwPVByxix7e/03P5PXg1OY27lBNZWpZJfXN48g3tZkfEt6uf3jaaxgGBj6OKQm6DreN8EBlF3bXsbj3EPw6k9RpDfsQA2vGXqEK5GaAyExzh8AHjYriixBvJDa43+CICgMOiUBqPvN4J54rCmMSqqBqRZxheKc+GnN2Hda8a48y6jYdr/QbcLa/VDtrJZKHv+pqN8teWoZT/PdOdeXum5Izw6zH1wdzVTFqCUEBZUjWJB2Si6q2yuC1zOlYGruCxwHfuqOhCUcStceIfRAdccHNtqdCJv/cz4Oh3fAy55FgbMkCF5TV2bnnDBQ8YDoLwESvKMeR4l5rke1WwXHrNuV7n+O7CIiDeCeNptxr8dBjb7ZrlGDe7H8jx8EtdU3mFjREBskvFVu7ZDuLxVeALWzYYNbxuBo8elMOaPxle2OoiwWbfVsc3cHKzdBeefHp3A8OeXAXCuvJJvdxwH7NMubDx4xmkpQFcydSLPVNzECxUz+GDkUQJ+fpe0jOdh00vQa5Ixc7LbhKb3R1B2DrZ/ARnvwJEM4y6s7xUw5Gbj/8Zv7ipamOAwCG5f885WraH8nOsPABVgdILGd/e734tGDe5nSz0HmBrJWg1f3mXdDww1pmLHJEJMZ9O/5keS8W9wLZoYcg/Cj68Y45+ryo0e/tF/gA4D6lT9u8Z140RBid2CHc8u3GnZbhsVSpHpPTuQU+T0/PjIENpGh/G3KwfQJT6Ca+asA4wJVObJVIfPnOPK19e6rcPDE3vzwjf2q0OVEkJJ6nRuWpfMSxcEMSL3KxKzFhlfmcPjoN+VRqDvNLRx/0BO7jQC+pZ5xmzehJ5G5+iAa5rPNw3he0pBSKTxiKn9ykbNjX91qPaZCu37Q3628cg7ZN3et9z4muY4yzOyjUPAT7Lfj0ywBqyTu4xO0q2fGp/4g66FUfdDvHfL9XliXtfVncjQIE4XlXEsv5jv91iXGQsPDqS4vJKhXVoDMH2YfUdPYWmFZXm/cX9faXduxrAk5m0wkn7uf34y87c4pg0ymIde/un7CmASPz74PB1zfoSt84x27A3/MWYwDrjG6HSKa6DVcspLjIRaGe/AoR8hMMT4PUi7zRjJ4Gd3Y0J4q1GDu8/nT4W2MoJ7ezdL9FWUGR2d5oCff9hoysnPNjpvMpcZX99sBYUZgT4s1viKHxwBI34L591Trwmafj24E19ssg+0qR2iOZBTxOP/s3awXjkkkeeu6Mf/Nh3hqqGu70ryz5Vbgntllf2bnhRnHZIZEKAICXTuWLx2eBK920cRHKgsbf1Ld5/h5vMnGgmxSvKNTq+tn8DK541H0kgjyPe9on7umnMyjbb0zXONfo7WKXDxM8aIF8ljIoSf3bl7EhRiTerkitZGW1z+YecPgMLjMPZBGHFXg2S7G9U9wSm4m9drXbbLmjr2pelGe/q1wzu7vVbuuTKX2SgB0kx3+2bDUoz9Z6b15UxRGSO7xjOyq/HzxkaEWLJV/nwo15pOOCwGhtxoPPIOwy+fGYF+4R/hm3Qjr82Aa6DnpRBUh0VTKspg90LIeNuYaakCofcU4y495QLwcnioEC1B4965N+aLu6KUcZcZEWf0ljeiqYM68qfPtgDw+JQ+9G4fzdnSCu78cKOlTN+OrhcXN/voNyO4/s315NmMmIkMCbRbUGSIQ3BvGxXmdhZq64hgS3Cfv/koM8d2pW/HGPtCsUlGp/LoPxhjhLd+agT7XV8b3376XmEE+pp0bOZmGUNMN31g5DiPSYILHzdyvDTVmYxCNLKWdefejJhXbErtEM1vxnQFjM5QW69fP7Taa7SLNjqLb377Jzb9+WKO5BXTJT6SHab0AvGRIS5XhnJnzwn7dWGnvLLGfToCpYzkWh0HGc0l+1cad/NbPzGaU2K7mNrnr4GE7s7Pr6yAvUuMu/TMZcb1elxq3KV3nyDj0oXwQIJ7E7bpzxdbJjSBcz4Zd00tZrER1tmwg/+y1LLdKTachyf1ZkRK3dvCS8orPS/kHRgEPS4yHqWFsPNrI8iv/jus+psxymbADOj3a2N278/vG4/CoxDVwRjnPOSmFjXSQYi6kuDehLV2WK/VYxB1EBvuOtXBkbxipg60riX7+Z3nWe7yqzOkcyw/OyzcXVhSUbN6hUYZo4wGXWssCP3LZ0bTzeIHYckjoKuMvo/uE2Dyi0aOFx+kSRWipfGv0TItwEMTe1nytnsS5GWTS1qyd3fwb9w4lKe/2sHCrccsxwpLyr3OUOkkugOMutd4nNhuBHoVaCRWa6ihlEL4qUYdXqCbXpdqk3fXBXUfU/+vGbVbaKJtVBizrxtid6zQi5muXmnXFy56Cib8WQK7ED4g33ebGaUU79w6jNCg2n8u17R5pzo+n2UshPAJGRjcDI3v1Zbzu9V+ok5QQN1mbY7qbh3nX+gqX7wQotFJcG+BOsfVbam4OTem8X/XDQbwKgGZEKLheRXclVITlVK7lVKZSql0F+dvUUqdUkptNj1+4811pUO1/i25fyz/mjGIl6dbJ2X1aBdVp2tGhgYxurvxzcFnbe5CCJ/y2OaulAoEZgMXA9nABqXUAq214yqwn2it76mHOoo66NU+il7tjWD+x0+3+Oy65pzz0iwjRNPkTYfqcCBTa70fQCk1D5gGVLPEt2iKvrx7FAdPO6cKro2gwAAiQgLrdOdeVlFFSB06hoUQ7nnzl9UJOGyzn2065uhKpdRWpdTnSimXiwsqpWYqpTKUUhnQBHPL+LlBSbFMG+S7TJZRYUG1vnOf8spqej6+mE2Hcn1WHyGEla9um74CkrXWA4ClwHuuCmmt52it07TWaaYDPnp50RhOFJTyaUZ2rQL89qNGfpu1+0/7ulpCCLwL7kcA2zvxRNMxC631aa11qWn3TaD6jFYm5lSyonnr/9S3NSq/fNcJy/bafRLchagP3gT3DUAPpVSKUioEmAEssC2glOpgszsV2IkH/Tq6orcVAAAgAElEQVTGcH53WVShJTqWX2LZXr03h93HCxuxNkL4J4/BXWtdAdwDLMEI2p9qrbcrpZ5RSk01FbtXKbVdKbUFuBe4xdN1ZfWz5u+aNJddKx6ZR9qYXfrPVb6ojhDChlfpB7TWi4BFDseesNl+BHjEt1UTTZ3tilDVKSmvpPefvyElIZIVD4wjwMUne2lFJaFBkqNdCF+RcWii1m45v4tlOyvH/RDLBZuPAnAgp4jiskpKK6oAePfWYZYyvR7/pp5qKUTLJMFd1NoVQ6yLZ1z7n3WW7aoqzctL93CqsJQXvtnFQ//dajn32spMHjAtH5jaofplAoUQtSfBXdRap9hwLurTFrDvJN1+tIBXlu3l3o838frKfXbPeeN7635ocKDdwt5r9ubUc42FaDkkuIs6mdy/g9Mx86xTV2PYyyutcxtCgwJob7MC1A1vrQfgux0n6PrIQkltIEQdSHAXdWJerq9vR2sTS1U1k9PibJYODA0K4M5xXe3Ol1dW8drKTKo07DwmQySFqC0J7qJOzEv5bT9awNdbjY7TikrXwT0oQHGmqMyyr5QiNCiQJ36Vajk2a/EuIk1DJYvKJOOkELUlwV34zD1zNwHGsEZXKqqsQf/924Zbtm3v9N9ac4DVprb3kjLX1xFCeCbBXdSZ7bD1jKwz5J3z3FY+2mZ28vRhridD3fXRz9LuLkQtSXAXdTbnxjTL9lVvrCU795zd+eeu6MebN6XZrfsaYLPUX3RYMFmzpri89rNfe8xkIYRwQYK7qLOLU9vZ7b/07R7AuDv/869SuX5EFy5Kbcf6RydUe52oMOcJ09uO5vuuokK0IF6lHxCiJgpLjY7Qd28dZulwBYiNCOFfMwYxIDHW5fNuPT+ZV5Zn2h3rVcclAYVoqeTOXfjEFIfx7oEByi6wm00b1ImUhEiX14iJCHE6drKw1EVJIYQnEtyFT/z96oFsefISy35lVc0XYokND3Y6tiYzp1bXMiurqOJEQYnngkL4GQnuwifCQwKJCQ+mq5u7cm/E2AT3rFlTLKNwDuScrfU1n5i/jRHPL6OkXIZVipZFgrvwqYjQ2qft7dQ63G7/zZuMUTjbjxaQnL6Q9TVYkk9rzbYj+Za0xDlnpXlHtCwS3IVPFZXW/g65j0OWyK5tWgHw3EJjOOQ7P2R5fa231hzgV6+u4ZSpzf6UtN2LFkZGywifiggx7txfvGpArZ5/SWo7kk1NOx1ijKRi5k5Vx3QEWTlFFJVV0LdjjNN1tmTbD6Fcf+AMgzu3rlWdhGiOJLgLnzJnhIxv5TzyxRtzbrJOiAoLDiQ6LIiCEiOop3a0v7Mf9/eVgNE+f7KghOW7TtK9bSvSkuNo0yrUruysxbu484JutaqTEM2RBHfhUyGm4Y++WjLPHNgBSsurLNtf/Jxt2U5OX2j3nA2PXcSP+yQ3vGjZpM1d+JT5zr2ssspDyZrLPVfGwq3HKK2o5EvT0n2u/G9TNruO26cLHtzZ9cQpIfyVV8FdKTVRKbVbKZWplEqvptyVSimtlEpzV0b4t8sGGPndU+JrPyTSnfmbj3L33J95+ds9dsMmHa3cfcrpWK5NqmFvzfvpEDe8ub7GzxOiKfAY3JVSgcBsYBKQClyrlEp1US4KuA+Qv4YWbPqwJLY9famlU7Su1jw8HoAEmzb8f6/a79Smbst2rZAHL+1Fj7atyPUiU6Wj9C9+YU1mDuX18C1EiPrmzZ37cCBTa71fa10GzAOmuSj3F+AFQKYDtnCtQn3XlZPYOoKsWVOcRsQcdsg8CfDJzJGAdXm/164fwt3juzO5fwcKSsqpqGWQ9iaFsRBNjTfBvRNw2GY/23TMQik1BEjSWtv3bAnhI5EOk6OW7jgBwH/vOt9ybETXeLsy5iyTrSOC0Rrum7fZ69fTNrf/eedq3qQjRGOr8y2WUioAeBm4xYuyM4GZAJ07d/ZQWgiryBDXv6rd2kSy5P6xnC11XpIv2DRyp32MMfN14S/HmO3l69mO0skvljt30fx4c+d+BLBdKifRdMwsCugHrFRKZQEjgQWuOlW11nO01mla67Q2bdrUvtaixYl009SjlKJX+yiGdjEmKN0w0nrTYB65MzwlrkavdfpsKUP/stSyL8FdNEfeBPcNQA+lVIpSKgSYASwwn9Ra52utE7TWyVrrZGAdMFVrnVEvNRYtUue4CABuOT/Z7rjNgk4APHhJb8u2ecx9XGSIJaFZVZUmv7ict9YcsGt6sfXtjhN2673e/l4G324/XtcfQYgG5TG4a60rgHuAJcBO4FOt9Xal1DNKqan1XUEhAK4cksjMsV354yU92fPsJMvxAGUf3aPDrXf4wTb55KcNMrqJuj66iMtn/8Bfvt7BZxuz2XmswOm1XKUYnvnBxjr/DEI0JK/a3LXWi4BFDseecFN2XN2rJYS9mIhgHp3cx+m4Y3BXSvHbsV3596r9tI4ItjluLXMgpwiAhz7fCuC0fqur9nshmhuZoSqatbBg51/hBy7txeqHxtM2OsxyrLDEfbv5OYeEZOUV9T+u/Xh+CcnpC/n5UG69v5ZomSS4i2bp69+PZvmfLkA53LmD0RyTZGqjN7vpvGS318rOLbbbL3UT3G3b6JPTFzrltKmJVXuNWbQfrD1Y62sIUR0J7qJZ6tcpxpLv3RtJcRGsenC8y3MnC+xzvdvmxbFNc1BQbNzh7z9lXRmqtmPgV+81EpsFOfYIC+EjEtxFi9E5PsLl8bxi+wA9Z9V+AEZ3T+DyQR0txwc+8y0HTxexJtOacfKCF1dSVYs1Xr/aYiQ+CwqU4C7qhwR30aLM+nV/p2Pu8s58+JsR9GwfZXfso/WHeGL+dst+fnE5f/zU+5mvgN0InSpJWyPqiQR30aKYm3JuG5ViOVZdxshrh3Xm8SnWUTq2TTJm1aUfdpRztpRJ/1ptfW1JbSDqiQR30aIMT4nj09+ex8OTelmGQL68dA9aa256+yfLZKVhycaM14AAxY3ndbE8/7udJ+v0+o5j6L815cgRwtdkJSbR4rhKR7DjWAGr9pxi1R5jFMuAROviHt6sKvXqsr18vfUY/RNj+PvVA92Wq3IxKzb/XDkxEe7z0wtRG3LnLlq0FFNagows+/HmmxzGn+98ZqLd/k+PTbDbf2npHnafKOTzjdlsPJjrdly9q77X4wWSJVv4ngR30aJ9/fvRgDWFsJnj3Xp4SCB3jzcW2A4JCqBtVBirH3I9tPLK13+k/1PfujxnO7ImPtJYgORofrHLskLUhQR30aJFhgYxKCnWbngjwFu3OK8U+eClvcmaNcWS2yYpLoK3XZSrjm2zTKVp+3h+CVk5RRzJkyAvfEfa3EWLt/lwnmX71lHJ3D46hQg3+eMddW8T5fZcWUWVJe2wmW2zTN65cgIUHMsvYdzfVwKw+L4x9OkQ7X3lhXBD7txFi3ffhB6W7REpcSS2dj3ZyRXbiVHT0xLtzrmavWp75/7Y5D60iQol+4x1yUDbYZJC1IUEd9HizRhuXYvmwt7tavz8LqYA73i3f8ZVcLe5db9+ZGfaR4fxxaYjTuWEqCsJ7qLFi48MtWw7NqN4Y+UD43jl2sH86ZKefHD7cMvxe+ZuItthIW9zbH/t+iFEhASxJTvf6Xq/+2ij24VEhPCWBHfR4oUEBfDWzWnMv3tUrZ6vlGLqwI5EhQUzpkcb5v5mBACZJ88y+oUVdmXNzTLV5Qtb9Mtxp0yVjkrKK3n26x3ku0mdIIQEdyGACX3aMTAp1nNBL/RoZ+1kDQm0/xMzz1A1pyoe38v1WsInC6sf+75w6zHeXHOAl5furktVhR+T4C6Ej9muAFVWWcW0/1tjWRDE3NoSaArus68fYilrO7P1SF71wX3t/tMA7M8p4tDpc9w/bxMfrpPc8MJKgrsQPhYUGMBNNvlotmTnc/HLqwCbZhnTX15ESBAx4cHcNa4bVw1N5LM7zwPgqIcx759vzAaMvPBjX1zBl5uP8viX26iolDSTwiDBXYh68My0fnb75glK5olLtitIbXnyEh6e2BuAYclxRIcFcayWE5pOnS31XEi0CF4Fd6XURKXUbqVUplIq3cX5O5VSvyilNiul1iilUn1fVSGal17t7Cc4JacvtIyCCXSxPKBZx9hwj80y7pwokOAuDB6Du1IqEJgNTAJSgWtdBO+5Wuv+WutBwN+Al31eUyGamY/uGMG/bxxqd6yi0jxaxn1w7xQb7rFZxrGj1uyEJCETJt7cuQ8HMrXW+7XWZcA8YJptAa11gc1uJCCDdEWLl9AqlEv7tmfvc5Msxz5cfwiofihkh9gwdhwrqDbXTJmbtvXFvxyrXWWF3/EmuHcCDtvsZ5uO2VFK3a2U2odx536vb6onRPMXHBjAP68ZZHcsoJro3jE2HIBRs5Z7/RqT+7cHjFWhavI84b981qGqtZ6tte4GPAw87qqMUmqmUipDKZVx6tQpX720EE3e5P4dUApLR2l1zTJto8Is29XNVL1xpHVEzmvXW5t/juQVczxfmmdaOm+C+xEgyWY/0XTMnXnA5a5OaK3naK3TtNZpbdq4nrwhhD8KCQpAa8g4aCwCEhzoPrh3axNp2V5pWhnq5aV7SP/vVgCWmJYC3J9zltYRwZYmnr4drdkkR/51GZsP51mGTJrlnC11mdBM+B9vgvsGoIdSKkUpFQLMABbYFlBK9bDZnQLs9V0VhfA/1eWwGdy5Nf+aYTTj3PrOBsorq3hl2V7mbThMRtYZfvvBRgCiw4L5MX0C256+FICF947hx/QLLde5fPYPPPDZFrtr3/DmegY9s5R3fzjg6x9JNDEeg7vWugK4B1gC7AQ+1VpvV0o9o5Saaip2j1Jqu1JqM/BH4OZ6q7EQzdT9F1nvgdyNdjGbOrCjZXvbEWtysaveWGu56395+iDCQwLtslF2iLE26ZjZLsq963ghAE99taOGtRfNjVdt7lrrRVrrnlrrblrr50zHntBaLzBt36e17qu1HqS1Hq+13l6flRaiOfr9hdbg3jbaOQjbsp3kdMVrP9qdKzcNpwwPcV64WylFWLD9n3VBcTn558rp9ugiu+PJ6Qs5W1rhXeVFsyMzVIVoIIEBigGJMQDEhAd7KA0X9Kxdv9S7tw632889V8a+nLN2d/Bm/Z5cUqvXEE2fBHchGtDcO0by06MTvCr79i3DavUaI7vG2+3nFZezak/9j07LPFlIRtaZen8d4R0J7kI0oFahQR6bZMwCAxT/+935lv1nL+/H2keMDtPHJvep9rljeiTQo20rwFju72Sh67QEKQmRLo9XZ9uRfJcjbmZ+sJGr3lhb4+uJ+iHBXYgmbHDn1pbt87rF0yEmnG1PX8pvxqRU+7wPbh/BR6ZFQ257N4O5ppmxF/Rsw9I/jLWUO5BTxKhZyznlJvg7WrH7JL96dQ2DnlnqdG7/qSKg+rH5ouFIcBeiiTMPi2xnuuNvFRpk1+HqjqtvCO/dNpxubVrZjdY5klfMl5uOcPu7G3jhm10cOn3O6XlmCzYftWz/+/t9LsvkyepQTYIEdyGauGmDOpE1awqtQoM8F3YwPS3R6VhAgGLPc5O4fbT17v/A6SKW7TrJ6yv3MfbFFZwpsja7HDxdxOwVmWit+Z/NYt5/XbzL5WtK2uGmQYK7EH7sxpHJlu0gh3w2j9q02zuOu995zJoL8J65m3hxyW5W782xKxMb4XrEz4+ZOS6Pi4YlwV0IP9bfNPQSoMJhKGRggLKs/PTuj1l25w6fMZpmLp/9A7+YJlHd9PZPlvMTercl71w5yekLWbDFaKrp2c7owH3qqx3kF0vTTGOT4C5EC/HbC7o6HRuWHOdyke69J88CsPlwnstrdY6PsGzf+/Embnt3A3tOnLUcG/j0t+w4WuDqqaKBSHAXws8lmwLxTecluzyf66ID9K01B/jBTfPKzLFdnVaZWr7rpFO5ya+slnHvjUiCuxB+buWD48maNYVOpjzxjq4c6tzpCnD9m+tdHtdaM22Q05IOLv1ikxdHNCwJ7kK0cKkdjFTB/TpFs+fZSZbx8WYTerdlz7OT+PB24/g943sQHhJI1qwpTtf6/sFx/PXX/S37rtZ0XbM3hxe+cT3SRvhOzcdWCSH8ypDOsdx0XhfuGNOVkKAABibF2p1vFxNGSFAAo3skuAzotjrHRbA/p8iyX1ZhXQ5wzqp9HMsv4Z0fsgB44JJeBFa33qCoE7lzF6KFU0rxzLR+JMUZbfOO4+kDq5kw9djkPoQEBVja4JVStLeZPHUs37oO7POLdlkCOxgLh3jj9ZX7uPntnyh3s26scE2CuxDCyRe/O5+ubSIZ1T3eLg+9ozvGdmXPs5N459ZhfP370QAktra27bsbbQNwosC6FGBG1hkm/nOVU+qCnw6c4YVvdvH9nlO8vtL1jFjhmmqsPBBpaWk6IyOjUV5bCFG/0v+7lXkbDgOQ0CqE/7tuCDPmrHMq98ik3gxMirWce+JXqYztmUByfCRBgQEkpy+0K++pWaglUEpt1FqneSwnwV0IUR+W7zrBbe9a/8YDFLhIKV8jEty9D+7SLCOEqBcX9m7HhN5tLftVGjrFhnudz7452X28sMllw5TgLoSoN/++cajd/oiUONpGh/G3qwbQtY13ueQHJcVy48guABSXVfq8jtuO5HMkr9hzQTd+PpTLpf9cxRPzt5ORdYaKJtLxK8FdCFFvghwSkgWb9qenJXHXBd3szrkbFbn5cB6DOxvDM4/m1z4Iu/OrV9cwatbyWj//4Glj6OcH6w5y1Rtr6f7YYl9VrU68Cu5KqYlKqd1KqUylVLqL839USu1QSm1VSi1TSnXxfVWFEM3RonvHWLaDg6wRvLtppSizt28ZZplF+9DEXnbn2scYwysnvPQ9uUXOq0DV1rky6wLh24/Wbjbty0v3+Ko6PuUxuCulAoHZwCQgFbhWKZXqUGwTkKa1HgB8DvzN1xUVQjRPqR2jGZ4SB1jv3ME5uI/r1ZaXpw8EYELvdpbjI7vGMchmYtXgvzivAgVQWlHJV1uOVtv2bTtWfv7mI/x1kXWm7JRX1pCcvpCCkppltJw+NMnp2AGbiVyNxZs79+FAptZ6v9a6DJgHTLMtoLVeobU2L9+yDnCdrEII0SKZl/Hbmm29O44KC2blA+Psyo3oGk/WrCn0ah9lWYHqjRuGEhHieTL9nR9s5Pcfb+IvX+90eX7z4Tx6PLaYabN/AOC+eZv5YN1Bp3LPL3T9fHfKXLSxv+eQQrkmHvniF0Y+v6zWzzfzJrh3Ag7b7GebjrlzO9A0Gp2EEE3C36827sgdm1SSEyKZe8cIFt472uk55hWoYiNCnM4lpy/kg3UHuf7Ndazff5p3fzjAit2nANjgJhPl5aagvuVwHrPcrCIFMG/DYapqMGbzjItmIsf8+K5sPpzHlFdWc9JmMhfAxz8d4nhBiV3qhtrwaYeqUuoGIA140c35mUqpDKVUxqlTp3z50kKIJqxvx2gGJsXyl8v7OZ07v1sCfTvGuHiWvT6mBGdmf/5yGz9knuaaOet46qsdluPxrZw/DBy9YbP+693ju3HXOPvO3d0nCj1ew6ygpIKkOOeMm8eq6fw9ll/M5bN/YPvRAu6bt9llmbfWHPC6Dq54E9yPALaNSommY3aUUhcBjwFTtdYuk0ZoredordO01mlt2jgvECCE8E9hwYHMv3sUo7on1Poa79wyjGem9fVYbuXuU7z3Y5al7X3F7pNOM11tlVVU8fDE3mTNmkJHU8dtTdrMC4rLiYsI4e9XD+SNG4ZYlh984LMtbp+z67j1w2Pt/tN257qY8u9X1XHcvDfBfQPQQymVopQKAWYAC2wLKKUGA//GCOzOWfuFEKKO2seEcdN5yXYpibvYrAhl68kF23l1eSYAt76zodrrjkiJt2wv+cNYwGj39samQ7l8v+cUW7LzuWpoIhP7deCHhy8E4IfM027v3h3Xs7VlHjH04pLdTHllNbMW7+L5RTspKq1w+xxXPAZ3rXUFcA+wBNgJfKq13q6UekYpNdVU7EWgFfCZUmqzUmqBm8sJIUSdjOqewAU9jW/+X/5ulNtyLy/dw4UvrbQ7duPILjw+xVgYfGzPNux+diIXpVpH5kSFGXfd+cXlVFZpXlm2t9ohkrZZLs0ibbJqnvfX5XbDLS31eOsnu/2zpRVk5RSRnL6QH/dZ7+S3Hy3gje/3MWfVfvo+uYTSCu8ncXmVz11rvQhY5HDsCZvti7x+RSGEqKO3bk7jdFEZrSND+OPFPd2ONd9/yr555empfVEKSiuquHxwJ0KDAt2+xs5jBby8dI/l2rv+MpGwYPvyg5JiWbDlKD89Zp9S4boRnZm7/hAAqU8sAWDlA+PYdjSfNXudly8c/tx3TpO6XNl93Pu+AFmsQwjR7AQFBtDOlDf+jjFdKauookNsGFFhwbSPDuNfy/bwQ6b1Dvjm87pw/0U9CTA1h9w9vrvba3/62/OY/u+1LN1xwu74X77ewXNXWFeZOni6iGe+NjpyY8KD7co+f0V/Fmw+ylmbppRxf1/p9FqrHhzP2BdXcK6skg0Hcz3+3LaLkHsi6QeEEM1aeEggD1zai+tHdGHqwI4MT4njToe74N+M6UrrSM+jaAAGJBojd/61bK/d8Y/WH2L/KWtwPXj6nGXb1TeAoV1aV/s6nWLD6WzTZ7Bqj3UEYUhQANufvtSyv+Exo3Gkuk5aRxLchRB+p38n69DKrFlTLKtMecOx6eW924Zbti986Xuvx5+/e+sw3r11GPPvHsWjk3s7nV90n5GWwZwUzdbIrvFEhgax8oFxPHVZKm2iQr2uv5kEdyGE34mNCOGre0az+L4xngu78MSvjAwr53eLZ2yPBL41jaIB6Pn4YgY98y3nTBkqzStQOVJKMa5XWwYmxXLTeclO583LGTqO/V9472hmXzcYMCZ53TIqBXCfWM0dCe5CCL/UPzHGaeKTt24bnULWrCnMvWMkSil6toviuSusQTjvXDnrTOPTI0Lcd8qaOX4bAOwWB19hSsPw0MRe9O0YYxm1Y+ub+8fy8R0jvf4ZZCUmIYTw0t4ThVz8j1V2x9Y/OsHSuVud0opK/rNqP9eN6ILWmvhWNW9qAVmJSQghfK5Huyinph5v7tzB6HS958IexEWG1Dqw14QEdyGEqIGe7aIs2z3atnLZhNIUSHAXQogaCAxQlk7QgTZ55psaCe5CCFFDF/RoQ4eYMKYO7NjYVXFLZqgKIUQNdY6PYO0jEzwXbERy5y6EEH5IgrsQQvghCe5CCOGHJLgLIYQfkuAuhBB+SIK7EEL4IQnuQgjhhyS4CyGEH5LgLoQQfsir4K6UmqiU2q2UylRKpbs4P1Yp9bNSqkIpdZXvqymEEKImPAZ3pVQgMBuYBKQC1yqlUh2KHQJuAeb6uoJCCCFqzpvcMsOBTK31fgCl1DxgGrDDXEBrnWU6593igkIIIeqVN80ynYDDNvvZpmNCCCGaqAbtUFVKzVRKZSilMk6dOtWQLy2EEC2KN8H9CJBks59oOlZjWus5Wus0rXVamzZtanMJIYQQXvAmuG8AeiilUpRSIcAMYEH9VksIIURdeAzuWusK4B5gCbAT+FRrvV0p9YxSaiqAUmqYUiobuBr4t1Jqe31WWgghRPW8WolJa70IWORw7Amb7Q0YzTVCCCGaAJmhKoQQfkiCuxBC+CEJ7kII4YckuAshhB+S4C6EEH5IgrsQQvghCe5CCOGHJLgLIYQfkuAuhBB+SIK7EEL4IQnuQgjhhyS4CyGEH5LgLoQQfkiCuxBC+CEJ7kII4YckuAshhB+S4C6EEH5IgrsQQvghCe5CCOGHJLgLIYQf8iq4K6UmKqV2K6UylVLpLs6HKqU+MZ1fr5RK9nVFhRBCeM9jcFdKBQKzgUlAKnCtUirVodjtQK7WujvwD+AFX1dUCCGE97y5cx8OZGqt92uty4B5wDSHMtOA90zbnwMTlFLKd9UUQghRE94E907AYZv9bNMxl2W01hVAPhDviwoKIYSouaCGfDGl1Exgpmm3VCm1rSFfvw4SgJzGrkQNNKf6Nqe6QvOqb3OqKzSv+jZmXbt4U8ib4H4ESLLZTzQdc1UmWykVBMQApx0vpLWeA8wBUEplaK3TvKlkY2tOdYXmVd/mVFdoXvVtTnWF5lXf5lBXb5plNgA9lFIpSqkQYAawwKHMAuBm0/ZVwHKttfZdNYUQQtSExzt3rXWFUuoeYAkQCLyttd6ulHoGyNBaLwDeAj5QSmUCZzA+AIQQQjQSr9rctdaLgEUOx56w2S4Brq7ha8+pYfnG1JzqCs2rvs2prtC86tuc6grNq75Nvq5KWk+EEML/SPoBIYTwQ40S3D2lM2gMSqkspdQvSqnNSqkM07E4pdRSpdRe07+tTceVUuoVU/23KqWG1HPd3lZKnbQdOlqbuimlbjaV36uUutnVa9VjfZ9SSh0xvb+blVKTbc49YqrvbqXUpTbH6/33RCmVpJRaoZTaoZTarpS6z3S8yb2/1dS1qb63YUqpn5RSW0z1fdp0PMWUpiRTGWlLQkzH3aYxcfdzNEBd31VKHbB5bweZjjf635lHWusGfWB0yu4DugIhwBYgtaHr4aJeWUCCw7G/Aemm7XTgBdP2ZGAxoICRwPp6rttYYAiwrbZ1A+KA/aZ/W5u2WzdgfZ8CHnBRNtX0OxAKpJh+NwIb6vcE6AAMMW1HAXtMdWpy7281dW2q760CWpm2g4H1pvfsU2CG6fgbwF2m7d8Bb5i2ZwCfVPdzNFBd3wWuclG+0f/OPD0a487dm3QGTYVtWoX3gMttjr+vDeuAWKVUh/qqhNZ6FcYopLrU7VJgqdb6jNY6F1gKTGzA+rozDZintS7VWh8AMjF+Rxrk90RrfUxr/bNpuxDYiTHjusm9v9XU1Z3Gfm+11vqsaTfY9NDAhRhpSsD5vXWVxsTdz9EQdXWn0f/OPGmM4O5NOoPGoIFvlVIblTGTFj3DcxAAAAKuSURBVKCd1vqYafs40M603RR+hprWrSnU+R7TV9i3zc0c1dSrwetragYYjHHX1qTfX4e6QhN9b5VSgUqpzcBJjEC3D8jTRpoSx9d2l8akQerrWFettfm9fc703v5DKRXqWFeHOjWFvzNAOlRtjdZaD8HIfnm3Umqs7UltfOdqkkOLmnLdbLwOdAMGAceAlxq3OvaUUq2A/wL3a60LbM81tffXRV2b7Hurta7UWg/CmNk+HOjdyFVyy7GuSql+wCMYdR6G0dTycCNWsUYaI7h7k86gwWmtj5j+PQn8D+MX8YS5ucX070lT8abwM9S0bo1aZ631CdMfTxXwH6xfqxu9vkqpYIxg+ZHW+gvT4Sb5/rqqa1N+b8201nnACuA8jCYM8xwb29e21EvZpzFp0Pra1HWiqSlMa61LgXdogu+tO40R3L1JZ9CglFKRSqko8zZwCbAN+7QKNwPzTdsLgJtMPeYjgXybr/ANpaZ1WwJcopRqbfrafonpWINw6JO4AuP9Ndd3hmmkRArQA/iJBvo9MbXpvgXs1Fq/bHOqyb2/7urahN/bNkqpWNN2OHAxRj/BCow0JeD83rpKY+Lu56jvuu6y+YBXGH0Dtu9tk/s7s9OQvbfmB0ZP8x6M9rfHGqMODvXpitEbvwXYbq4TRnvfMmAv8B0Qp60967NN9f8FSKvn+n2M8XW7HKMN7/ba1A24DaMzKhO4tYHr+4GpPlsx/jA62JR/zFTf3cCkhvw9AUZjNLlsBTabHpOb4vtbTV2b6ns7ANhkqtc24Ambv7efTO/TZ0Co6XiYaT/TdL6rp5+jAeq63PTebgM+xDqiptH/zjw9ZIaqEEL4IelQFUIIPyTBXQgh/JAEdyGE8EMS3IUQwg9JcBdCCD8kwV0IIfyQBHchhPBDEtyFEMIP/T/vMSRwXXQANAAAAABJRU5ErkJggg==\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.8338926434516907.\n", "Epoch 4: reducing lr to 1.9887481084816066e-05\n", "Better model found at epoch 8 with accuracy value: 0.8361297249794006.\n", "Better model found at epoch 9 with accuracy value: 0.8428411483764648.\n", "Better model found at epoch 10 with accuracy value: 0.8450782895088196.\n", "Better model found at epoch 11 with accuracy value: 0.8478747010231018.\n", "Better model found at epoch 12 with accuracy value: 0.853467583656311.\n" ] } ], "source": [ "learn.fit_one_cycle(epochs_all, max_lr = slice(1e-5, 1e-4), wd=0.05, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_stage2_resnet50_fold2'),\n", " ReduceLROnPlateauCallback(learn, monitor='accuracy', mode='max', patience = patience)])" ] }, { "cell_type": "code", "execution_count": 47, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "1788\n" ] }, { "data": { "text/plain": [ "True" ] }, "execution_count": 47, "metadata": {}, "output_type": "execute_result" } ], "source": [ "learn.load('Best_model_sz224_stage2_resnet50_fold2')\n", "\n", "interp2 = ClassificationInterpretation.from_learner(learn)\n", "print(interp2.confusion_matrix().sum())\n", "\n", "losses,idxs = interp2.top_losses()\n", "len(data.valid_ds)==len(losses)==len(idxs)\n" ] }, { "cell_type": "code", "execution_count": 48, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.6378821047093878" ] }, "execution_count": 48, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix2 = interp2.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix2.sum(axis=0) - np.diag(confusion_matrix2) \n", "FN = confusion_matrix2.sum(axis=1) - np.diag(confusion_matrix2)\n", "TP = np.diag(confusion_matrix2)\n", "TN = confusion_matrix2.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "\n", "TPR.mean()" ] }, { "cell_type": "code", "execution_count": 49, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 00:42

" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "0.6447937710613377" ] }, "execution_count": 49, "metadata": {}, "output_type": "execute_result" } ], "source": [ "interp2_tta = ClassificationInterpretation.from_learner(learn, tta=True)\n", "\n", "confusion_matrix2_tta = interp2_tta.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix2_tta.sum(axis=0) - np.diag(confusion_matrix2_tta) \n", "FN = confusion_matrix2_tta.sum(axis=1) - np.diag(confusion_matrix2_tta)\n", "TP = np.diag(confusion_matrix2_tta)\n", "TN = confusion_matrix2_tta.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "\n", "TPR.mean()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold 3" ] }, { "cell_type": "code", "execution_count": 50, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "this Learner object self-destroyed - it still exists, but no longer usable\n" ] } ], "source": [ "gc.collect()\n", "learn.destroy()\n", "seed_everything(SEED)\n", "bs = 32\n", "data = get_data(metadata_train_valid3, bs, 224)\n", "\n", "#Define Convolutional Neural Network with backbone resnet50\n", "learn = cnn_learner(data, models.resnet50, metrics = [error_rate, accuracy],\n", " callback_fns=[ShowGraph])\n", "\n", "learn.loss_func=nn.CrossEntropyLoss(weight=class_weights)\n" ] }, { "cell_type": "code", "execution_count": 51, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Learn rate finder\n", "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 52, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 02:54

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
01.5908840.8705340.2936150.70638500:43
11.0256910.4600510.1579250.84207500:43
20.8196320.4418240.1493730.85062700:43
30.6941790.4445190.1539340.84606600:42
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.7063854336738586.\n", "Better model found at epoch 1 with accuracy value: 0.8420752286911011.\n", "Better model found at epoch 2 with accuracy value: 0.8506271243095398.\n" ] } ], "source": [ "#Fit heads\n", "learn.clip_grad(1.0);\n", "lr= 1e-3\n", "learn.fit_one_cycle(epochs_head, max_lr = lr, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_resnet50_fold3')])" ] }, { "cell_type": "code", "execution_count": 53, "metadata": {}, "outputs": [], "source": [ "learn.load('Best_model_sz224_resnet50_fold3')\n", "\n", "learn.unfreeze()" ] }, { "cell_type": "code", "execution_count": 54, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 55, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 14:41

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
00.7179230.4376750.1505130.84948700:58
10.7080050.4335330.1488030.85119700:57
20.6101810.4088900.1294180.87058200:58
30.5874480.4327120.1465220.85347800:57
40.5970280.4296400.1396810.86031900:58
50.4951780.4509950.1374000.86260000:57
60.4477680.4582310.1362600.86374000:57
70.3397800.4650340.1351200.86488000:59
80.3065430.4589700.1402510.85974900:59
90.2546530.5190760.1442420.85575800:59
100.2782680.4759270.1311290.86887100:58
110.2312850.4441550.1271380.87286200:58
120.2094460.4654560.1305590.86944100:58
130.2139500.4612540.1277080.87229200:58
140.1834320.4631500.1271380.87286200:59
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.8494868874549866.\n", "Better model found at epoch 1 with accuracy value: 0.8511972427368164.\n", "Better model found at epoch 2 with accuracy value: 0.8705815076828003.\n", "Epoch 6: reducing lr to 1.7330529396223397e-05\n", "Epoch 10: reducing lr to 6.346644369977027e-06\n", "Better model found at epoch 11 with accuracy value: 0.8728620409965515.\n" ] } ], "source": [ "learn.fit_one_cycle(epochs_all, max_lr = slice(1e-5, 1e-4), \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_stage2_resnet50_fold3'),\n", " ReduceLROnPlateauCallback(learn, monitor='accuracy', mode='max', patience = patience)])\n" ] }, { "cell_type": "code", "execution_count": 56, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "1754\n" ] }, { "data": { "text/plain": [ "True" ] }, "execution_count": 56, "metadata": {}, "output_type": "execute_result" } ], "source": [ "learn.load('Best_model_sz224_stage2_resnet50_fold3')\n", "\n", "interp3 = ClassificationInterpretation.from_learner(learn)\n", "print(interp3.confusion_matrix().sum())\n", "\n", "losses,idxs = interp3.top_losses()\n", "len(data.valid_ds)==len(losses)==len(idxs)\n", "\n" ] }, { "cell_type": "code", "execution_count": 57, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.7765075886881279" ] }, "execution_count": 57, "metadata": {}, "output_type": "execute_result" } ], "source": [ "\n", "confusion_matrix3 = interp3.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix3.sum(axis=0) - np.diag(confusion_matrix3) \n", "FN = confusion_matrix3.sum(axis=1) - np.diag(confusion_matrix3)\n", "TP = np.diag(confusion_matrix3)\n", "TN = confusion_matrix3.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "\n", "TPR.mean()" ] }, { "cell_type": "code", "execution_count": 58, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 00:41

" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "0.7898711317061807" ] }, "execution_count": 58, "metadata": {}, "output_type": "execute_result" } ], "source": [ "interp3_tta = ClassificationInterpretation.from_learner(learn, tta=True)\n", "\n", "\n", "confusion_matrix3_tta = interp3_tta.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix3_tta.sum(axis=0) - np.diag(confusion_matrix3_tta) \n", "FN = confusion_matrix3_tta.sum(axis=1) - np.diag(confusion_matrix3_tta)\n", "TP = np.diag(confusion_matrix3_tta)\n", "TN = confusion_matrix3_tta.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "\n", "TPR.mean()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold 4" ] }, { "cell_type": "code", "execution_count": 59, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "this Learner object self-destroyed - it still exists, but no longer usable\n" ] } ], "source": [ "gc.collect()\n", "learn.destroy()\n", "seed_everything(SEED)\n", "bs = 32\n", "data = get_data(metadata_train_valid4, bs, 224)\n", "\n", "#Define Convolutional Neural Network with backbone resnet50\n", "learn = cnn_learner(data, models.resnet50, metrics = [error_rate, accuracy],\n", " callback_fns=[ShowGraph])\n", "\n", "learn.loss_func=nn.CrossEntropyLoss(weight=class_weights)\n" ] }, { "cell_type": "code", "execution_count": 60, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Learn rate finder\n", "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 61, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 02:53

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
01.4926282.0184860.6011400.39886000:42
11.0484031.9890570.6170130.38298700:43
20.8125971.5554020.5494510.45054900:44
30.6714371.7382900.5750920.42490800:42
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.3988603949546814.\n", "Better model found at epoch 2 with accuracy value: 0.450549453496933.\n" ] } ], "source": [ "#Fit heads\n", "learn.clip_grad(1.0);\n", "lr = 1e-3\n", "learn.fit_one_cycle(epochs_head, max_lr = lr, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_resnet50_fold4')])" ] }, { "cell_type": "code", "execution_count": 62, "metadata": {}, "outputs": [], "source": [ "learn.load('Best_model_sz224_resnet50_fold4')\n", "\n", "learn.unfreeze()" ] }, { "cell_type": "code", "execution_count": 63, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Learn rate finder\n", "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 64, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 14:11

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
00.7106401.5621700.5494510.45054900:55
10.6864821.6282190.5437530.45624700:56
20.6607321.8055000.5848600.41514000:56
30.5930291.5816860.5111930.48880700:56
40.5307951.4923370.5266590.47334100:56
50.4319301.6915960.5413110.45868900:57
60.4137621.5219120.4831090.51689100:55
70.4008821.8389670.5160770.48392300:56
80.3175341.3741710.4424090.55759100:56
90.2705731.4840090.4550260.54497400:57
100.2235831.2502410.4000810.59991900:56
110.2157171.3717410.4289780.57102200:56
120.1893991.3028130.3976390.60236100:56
130.1922041.5148610.4383390.56166100:55
140.1844561.3324000.4078140.59218600:56
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.450549453496933.\n", "Better model found at epoch 1 with accuracy value: 0.4562474489212036.\n", "Better model found at epoch 3 with accuracy value: 0.4888074994087219.\n", "Better model found at epoch 6 with accuracy value: 0.5168905258178711.\n", "Better model found at epoch 8 with accuracy value: 0.5575905442237854.\n", "Better model found at epoch 10 with accuracy value: 0.5999186038970947.\n", "Better model found at epoch 12 with accuracy value: 0.6023606061935425.\n" ] } ], "source": [ "learn.fit_one_cycle(epochs_all, max_lr = slice(1e-5, 1e-4), wd=0.05, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_stage2_resnet50_fold4'),\n", " ReduceLROnPlateauCallback(learn, monitor='accuracy', mode='max', patience = patience)])\n" ] }, { "cell_type": "code", "execution_count": 65, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "2457\n" ] }, { "data": { "text/plain": [ "True" ] }, "execution_count": 65, "metadata": {}, "output_type": "execute_result" } ], "source": [ "learn.load('Best_model_sz224_stage2_resnet50_fold4')\n", "\n", "interp4 = ClassificationInterpretation.from_learner(learn)\n", "print(interp4.confusion_matrix().sum())\n", "\n", "losses,idxs = interp4.top_losses()\n", "len(data.valid_ds)==len(losses)==len(idxs)\n", "\n", "\n" ] }, { "cell_type": "code", "execution_count": 66, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.7053206236926929" ] }, "execution_count": 66, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix4 = interp4.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix4.sum(axis=0) - np.diag(confusion_matrix4) \n", "FN = confusion_matrix4.sum(axis=1) - np.diag(confusion_matrix4)\n", "TP = np.diag(confusion_matrix4)\n", "TN = confusion_matrix4.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "TPR.mean()" ] }, { "cell_type": "code", "execution_count": 67, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 00:58

" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "0.7151314019463104" ] }, "execution_count": 67, "metadata": {}, "output_type": "execute_result" } ], "source": [ "interp4_tta = ClassificationInterpretation.from_learner(learn, tta=True)\n", "confusion_matrix4_tta = interp4_tta.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix4_tta.sum(axis=0) - np.diag(confusion_matrix4_tta) \n", "FN = confusion_matrix4_tta.sum(axis=1) - np.diag(confusion_matrix4_tta)\n", "TP = np.diag(confusion_matrix4_tta)\n", "TN = confusion_matrix4_tta.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "TPR.mean()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Fold 5" ] }, { "cell_type": "code", "execution_count": 68, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "this Learner object self-destroyed - it still exists, but no longer usable\n" ] } ], "source": [ "gc.collect()\n", "learn.destroy()\n", "seed_everything(SEED)\n", "bs = 32\n", "data = get_data(metadata_train_valid5, bs, 224)\n", "\n", "#Define Convolutional Neural Network with backbone resnet50\n", "learn = cnn_learner(data, models.resnet50, metrics = [error_rate, accuracy],\n", " callback_fns=[ShowGraph])\n", "\n", "learn.loss_func=nn.CrossEntropyLoss(weight=class_weights)\n" ] }, { "cell_type": "code", "execution_count": 69, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Learn rate finder\n", "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 70, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 02:58

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
01.4934033.4748340.7972540.20274600:44
11.0323102.0294260.7148740.28512600:43
20.8348952.0033900.6695650.33043500:45
30.6132291.9238280.6723110.32768900:43
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "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\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.20274598896503448.\n", "Better model found at epoch 1 with accuracy value: 0.28512585163116455.\n", "Better model found at epoch 2 with accuracy value: 0.33043476939201355.\n" ] } ], "source": [ "#Fit heads\n", "learn.clip_grad(1.0);\n", "lr = 1e-3\n", "learn.fit_one_cycle(epochs_head, wd=0.05, max_lr = lr, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_resnet50_fold5')])\n" ] }, { "cell_type": "code", "execution_count": 71, "metadata": {}, "outputs": [], "source": [ "learn.load('Best_model_sz224_resnet50_fold5')\n", "\n", "learn.unfreeze()" ] }, { "cell_type": "code", "execution_count": 72, "metadata": {}, "outputs": [ { "data": { "text/html": [], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "#Learn rate finder\n", "learn.lr_find()\n", "learn.recorder.plot()" ] }, { "cell_type": "code", "execution_count": 73, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 14:42

\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
epochtrain_lossvalid_losserror_rateaccuracytime
00.5997511.8980710.6604120.33958800:56
10.6128041.9147530.6746000.32540000:57
20.6240662.1382070.6864990.31350100:57
30.5515732.3060920.7290620.27093800:58
40.5228372.1330910.6402750.35972500:58
50.4468361.8791090.6137300.38627000:57
60.3766462.0465310.6201370.37986300:57
70.3544992.1684310.6530890.34691100:57
80.3122101.9445730.5913040.40869600:58
90.2471571.9181370.5546910.44530900:57
100.2062321.9094550.5606410.43935900:59
110.2024862.0228980.5473680.45263201:01
120.1823511.9766550.5446220.45537801:02
130.1595201.8155520.5135010.48649900:58
140.1646801.9989000.5510300.44897000:58
\n" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4wLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvqOYd8AAAIABJREFUeJzt3Xd4VFX6wPHvSScNQuihhN5774h0EBQRULGggiIuqz/dXeyoqOhadl0VxIKKoCKoiIAUpUsxSJFeA4QWEkgIpCfn98eZNNImZJKZXN7P8+TJzJ07974ZhnfunPIepbVGCCGEtbg5OwAhhBCOJ8ldCCEsSJK7EEJYkCR3IYSwIEnuQghhQZLchRDCgiS5CyGEBUlyF0IIC5LkLoQQFuThrBNXqlRJh4aGOuv01paSAJfCITURPH0hKBQ8vJ0dlRDCAbZv3x6lta5c2H5OS+6hoaGEhYU56/TWlJ4OWz6AX1+GcjWhyyOw6b+QdhGGvg2txzo7QiFEMSmlTtizn9OSu3Cwy2fgh0fg+DpoMgxueQ/8gqHlHbBoAvzwMBxdA0PfAu8AZ0crhChh0uZuBfsWw8xuEPGHSepjvjKJHaB8TbhvCfR5Gv5aAB/1hjM7nBuvEKLESXIvy5KuwOLJsOBe067+8AZofx8olXM/dw/oMxXu+9m0x3/SHzZ/AFIRVAjLkmaZsipiO3z/EFw8Dj2fMsnb3bPg54R2h0mbzAfCimfg2Fq4dSb4VSqVkIUorpSUFCIiIkhMTHR2KCXOx8eHmjVr4ulZyP/rfEhyL2vS02DDO7D2dQisAeOXQZ1u9j/ftyKMnQ/bPoaVz8HM7jByNtTrXXIxC+EgERERBAQEEBoairr2G6qFaK2Jjo4mIiKCunXrXtcxpFmmLLl0AuYMgTXTocVIeGRj0RJ7BqWg80SY8KvpXP1yhBlhk5bq+JiFcKDExESCg4MtndgBlFIEBwcX6xuKJPeyYvcCmNUDIvfByI/h9k+gXIXiHbNaS3h4HbS9Gza8DZ8PgZiTjolXiBJi9cSeobh/pyR3V5cQAwsfhO8nQNXm5mq91WjHHd/LD0Z8ALd/Cuf3mQ+QfYsdd3whhFNIcndl4ZtMst37A/R9Du5fCkF1SuZcLUfBIxsguIEZfbPkcTOyRgiRKSYmhg8//LDIzxsyZAgxMTElEFH+JLm7orQU0wb++VAzAubBVdDrH+DmXrLnrVgXxv8C3f8O2+fA7Jsgcn/JnlOIMiS/5J6aWnB/1bJly6hQoZjNqEUkyd3VRB2BT/ubNvC248zY9ZrtS+/8Hl7Q/2UY9z3ER8HsPhD2mYyJz0taKhxYCnHnnB2JKCVTp07l6NGjtGnTho4dO9KzZ0+GDx9Os2bNALj11ltp3749zZs3Z/bs2ZnPCw0NJSoqivDwcJo2bcqECRNo3rw5AwYMICGhZL4hy1BIV6E1/PkF/PK0KfI1ei40G+68eBrcDJN+N2ULfn7CjIm/5b9QLsh5MbmK9HTY+70Zjhp9BMrXgnsXQ3B9Z0d2Q3lpyV72nbns0GM2qxHIi7c0z/fxGTNmsGfPHnbu3MnatWsZOnQoe/bsyRyu+Nlnn1GxYkUSEhLo2LEjt99+O8HBwTmOcfjwYb7++ms+/vhjRo8ezaJFixg3bpxD/w6QK3fXcDUavh0HS/4OtTqZpOrMxJ7BvwrcvchcyR9YCrN6wsmtzo7KebSG/T/DrO6w6EFw94Ihb0FKPMwZbDqkxQ2lU6dOOcahv/fee7Ru3ZouXbpw6tQpDh8+nOs5devWpU2bNgC0b9+e8PDwEolNrtydLXwTLHwAEi7CgFehy6Pg5kKfuW5upg2+Tg9Y9IBJYjc9Az2eKPk+AFehNRxZDb9Nh7M7Tafz7Z9C85Hm9anby8wV+HwIjFsEIaXYjHYDK+gKu7T4+fll3l67di2rV69m8+bN+Pr60qdPnzzHqXt7Z5Xfdnd3L7FmGRfKIjegg8th7m1mItGE36DbY66V2LOr2R4eXg/Nb4XfXoFFD5nZslZ3fD18NhDmjTIfwCM+hEe3mtFFGf9WlRvDA7+AdyB8MRzCNzo3ZlFiAgICiIuLy/Ox2NhYgoKC8PX15cCBA2zZsqWUo8tJrtyd5a+Fpj27Witztedb0dkRFc6nvLlirdYSVk8Dn0AY9p/chcqs4ORWMxP4+HoIqAFD34G295gO57wEhZoE/+Wt8NXtpjJnw/6lGrIoecHBwXTv3p0WLVpQrlw5qlatmvnYoEGDmDVrFk2bNqVx48Z06dLFiZGC0k4aBdGhQwd9wy7Wsf1zM468Tje48xuTJMua1S/BxndM80y/ac6OxnHO7IA1r8HhleBXGXo+Ce3Hg6ePfc+/Gg1f3Wba32//GJrfVrLx3mD2799P06ZNnR1Gqcnr71VKbddadyjsuXLlXtp+/58p2NWgP4z+Erx8nR3R9bn5BUiMgY3vgk8F6PG4syMqnvP7YM2rcOBn8/f0mwadJpoZvEXhF2zq588bbfpSkq+aIa1ClDJJ7qVFazN0bt0b0OxWUx8mv6/4ZYFSZqRIYiysftHUuWl/v7OjKrqoI+bfZc8i0/fR52noMsk0QV0vn/Jwz/dmBNTiyabufpdHHBezEHaQ5F4atDb107d8aK7ibnnPGiNN3Nzhto8gKc40M3kHmmqVZcGlE7DuTdg1Hzx8TPNSt785ru/Dy880uS18AH75FyTHmbr7VuyfEC5JkntJS08z49d3zIXOk2Dga647IuZ6uHvCHV+YTsTvJ5oE37Cfs6PK3+UzsP4t+PNLUG7m36THE+Bf6GLyRefhbV6bxZPNMMqkOOj3kiR4USokuZek1GT4YaIp/NX7X+YrvxX/Y3v5wl3fmFo4346De3+E2s4dKZDLlUjTP/DHp6DTzXKEPZ80C56UJHcPs9qVlx9s+q9J8EPettYHvHBJktxLSkqCqa54eCX0fwW6T3F2RCXLpzyM+wHmDDKdiff/DNVbOTsq0yS27WPTL5CaBG3uhF7/LLnqmnlxc4Ohb5s2/U3/MZ2sIz40iV+IEiKXDyUh8TJ8NQoOrzLjwK2e2DP4V4Z7fjRJ7KuREH3UufHEnYd5d8Dyf0Cd7jB5m6ldX5qJPYNS0P8lM8po97fw3X3mw0ZYmr+/PwBnzpxh1KhRee7Tp08fSmJYeKHJXSlVSym1Rim1Tym1Vyn19zz2UUqp95RSR5RSu5VS7RweaVkRf9FMRT+52ayW1GG8syMqXRVqmWYZnW5eh9jTzonj4HKY2RXCN5hRPXd/B5UaOCeW7Ho+CYP/bYZczh9jruJF/q5Gw5rXTb/F1tnmA7EMzoyuUaMGCxcuLNVz2vO9MBV4Umv9p1IqANiulFqltc5eJWkw0ND20xmYaft9Y4k7Z8oJRB+FsfOg8WBnR+QclRqaksFf3AJzb4Xxy8GvUumcO/kqrHjW1KOv1hJGfgJVmpTOue3V2TZ+/qfHYO5IuOvb4i+ZaDWXz8Dv75t/x5R4U410x1cwcAGc02aEk2c58PS1/S4HbiXfzDV16lRq1arF5MmTAZg2bRoeHh6sWbOGS5cukZKSwvTp0xkxYkSO54WHhzNs2DD27NlDQkIC48ePZ9euXTRp0sR5JX+11meBs7bbcUqp/UAIkD25jwC+1Ga66xalVAWlVHXbc28MMSfNlWrceXOVWK+3syNyrhptzFDAr0aakTT3LSn5mbhndpiaN9FHodsUs3qVh3fhz3OGtnebBL/oIfMheM8PpfcB6MouHjf9Ejvnmyv0lneY0UyVG8Pl03AiEvyrmYT/23SIOpT1XOVm+3E3v93cgSIOYKjWEgbPyPfhMWPG8Pjjj2cm9wULFrBixQqmTJlCYGAgUVFRdOnSheHDh+e7BurMmTPx9fVl//797N69m3btitDQcXy93bsW6aNOKRUKtAWurfsaApzKdj/Cti1HcldKTQQmAtSuXbsop3ZtUYdNYk++Yup61+ro7IhcQ2h3Mwv3m7vg6zth3EJzheVo6WkmIax5Dfyrwn0/mUqNrq75rSbBfzsO5gwxzVklPXrHVUXuN6OZ/lpoknLbcaYaaVBo1j7la4JnHARWN/d9K4HXaVNfX6eDTjO/07OtipQr4btRnK7Gtm3bEhkZyZkzZ7hw4QJBQUFUq1aNJ554gvXr1+Pm5sbp06c5f/481apVy/MY69evZ8oU0w/XqlUrWrWyY+CB1uY9/uvLdsdqd3JXSvkDi4DHtdbXVSFfaz0bmA2mtsz1HMPlnN1tmmKUMmucVmvp7IhcS6OBZqLToofgu/tNQS13T8cdP+Yk/PAInNhk6rgMe7dsLSjSsL9pwpo/Bj4bZC4OKtYt/HlWcfpPs+rYgZ/B08/MDu76WFYCL0h+V9hpqZCaYK7uk22/07J1Xrt5ZDXnePmbAQBFGKJ8xx13sHDhQs6dO8eYMWOYN28eFy5cYPv27Xh6ehIaGppnqd/rlhADPz4KB5faahV9YdfT7EruSilPTGKfp7X+Po9dTgO1st2vadtmbae2mVEx3v7mP2Wlhs6OyDW1HGXq0Cx90rxJb/vIMeO8d39njqnT4dZZ0Hps2ZxHENod7ltsmq/mDDYjjlytn8DRwjfBhrfg6G9mGG2vf5rE7ogZwu4e4B5gknaG9DQzPDklPut30mXgvJl4V6GWWXzFDmPGjGHChAlERUWxbt06FixYQJUqVfD09GTNmjWcOHGiwOf36tWL+fPn07dvX/bs2cPu3bvz3zkt2Sx1GXsKBs2Azo/gsOSuTMPRp8B+rfU7+ez2E/CYUuobTEdqrOXb24+tha/vgoCqJrFXsFAzU0no+JC5AvntFfOfeci/rz8RJ8TAsqfgr++gVmcYOTvn1/eyKKQ93L/MdEB/PsRczddo4+yoHCtj0ZMNb5vRZH6VTYG2Dg+WfH+Mm7u5CPP2z9qWng7x0abzNvKAafYpF1To+7J58+bExcUREhJC9erVufvuu7nlllto2bIlHTp0oEmTgj+YJ02axPjx42natClNmzalfft8FneJj4Yr5yE10bQKFHFiYKElf5VSPYANwF9Aum3zM0BtAK31LNsHwPvAICAeGK+1LnDgZpku+XtgmRmnHNzQdIQFVC38OcL85175HGx+31yp9X226McI32Tq4F8+A32mQo//s9ZkoOijpiZ8YoxZs7ZaSwgMKbvVQ8Ek0QNLTFI/uwsCa5r29Hb3FLkPpkRK/qYkQswJczXvU96sievIpsOi0ukQGwHx0ew/fdn8vf5VMh92WMlfrfVGCulyto2SmWxH2GXf7u9McqnRBu5eWDYW2XAVSsGA6aaS5Po3zfC/rna+bVKTTfXGje+aq/QHV0LNQt/fZU9wfXhguemgX5htjoRPBXNlGRhiOl3Lh9huh2Rtc7UPgLQU00G68R0zqqVifTOJrOVo16qI6ukDlRqZEhVxZyH5gEnwzhiempoEl8LNB41/FfDzzpHYi8JClzylIGwO/PwEhPaAO7/O2aYn7KOUuSJNumwqZXoHmiu4gkQdNh2yZ3ea1ZAGzcj59dpqytc0Sxqe2WEmgV22/WTcPh1mvrJfq1xFk+jL25J9YIjtA6FG1gdASYxWulZKIuycZ0Z3xJyEqi1g1Gem1LWrVkNVynwD9wk0FUMvHYfEiua1LIXx84CZ2X4p3NwOqms+XNR1jV0BJLnnpLWZBJNwKfdP5H7Y9hE0HAijvyid/yRW5eZu6tknxcGSKearcLPhuffT2kxiWfGsGa8+em7e+1mRl5+5iMhPSoJpmrp82vyOjci6H3vadPYnXMz9PN9gCKhuGy3iY357+GTd97BNCCrwMduPxzX76TSzytjv78OVc1Czo5mN22igQzu6tdb5jiEvNs9yULmRma9y5Zx5j1aoXbJ9Alqbc8WdM69jxbrg4UNxV8mzZnJPTzdXhnkl6YSYfLbbftJT8j9uqzEw/H3X+kpZVnl4m2GRX94Kix4E7wVQ/6asx69GwU9/g4PLoN5NprKiPcPjbhSe5UwTTnD9/PdJjjfNDJmJP8Ik/ivnbaNGEk0TWUqCuZ0SbzrvUuJNu+/1qtvbLDEY2tPho5d8fHyIjo4mODi45BK8cjPvNZ9A0xZ/8agZUx9Yw/HfPNJSISbcfIiUq2i+abm5o7UmOjoaHx87l3fMg7XWUD21zYwXTowp+M3p5W96xctVsP2280eu1h0v4RLMGWq+jmZMADu8ygyZTIwx9c87PyIlckuT1qa9PDXBlvgTspJ+SqLtfh4fCGkpUL9vifaFpKSkEBER4dhx5AXR2nwAJl0GN0/Tx+aoWc+pyRAfZYZplgvK1dTo4+NDzZo18fTM2bl7Y66h6l/VrARUUIL2qSBX3q6kXJBZku6zQTBvFDQdZmqIVGlmRiJVa+HsCG88Spn/Ix5exVtusAR4enpSt64TJnmFb4IfJ5k+hG6PwU3P2b9o+rW0Ns1Xy/9pctYdX0DNfIZDFoO1rtxF2XUp3CT4uLNmdaR+067/P48QJSHpCqx6HsI+g8pN4LZZUKNt0Y6RkmAm3u2cZ77ljPzELKpeBPZeuUtyF64jNsK0B4c4/ipGCIc5shoW/828V3v9A3o9Zd+4+IvH4Nt74fxfZmW23v+6rjZ8e5O7NGQK11G+piR24foa9INHfzcVK9fNgI/7wvl9BT/n4HL4qI8pI3DXd3DTMyU+LFSSuxBCFFW5IBj5kRmee/kMzO4NG/+TeyGRtFRY/RJ8PdYMcXx4PTQaUCohSnIXQojr1Ww4PLoFGg4w6/TOGZy1vOSVC/DVbWaGbrv74IEVpbrEo7VGywghRGnzr2zmbOxeAMv+AbN6mLLFO74yE8lGfGDq05cySe5CCFFcSkHrMWZW8U9/M7WTgkLhwVVQ3Y7FOEqAJHchhHCU8iEwbpEpCV6jrVPXxpXkLoQQjqRUzlIaTiIdqkIIYUGS3IUQwoIkuQshhAVJchdCCAuS5C6EEBYkyV0IISxIkrsQQliQJHchhLAgSe5CCGFBktyFEMKCJLkLIYQFSXIXQggLkuQuhBAWJMldCCEsSJK7EEJYkCR3IYSwIEnuQghhQZLchRDCgiS5CyGEBUlyF0IIC5LkLoQQFiTJXQghLEiSuxBCWJAkdyGEsKBCk7tS6jOlVKRSak8+j/dRSsUqpXbafl5wfJhCCCGKwsOOfT4H3ge+LGCfDVrrYQ6JSAghRLEVeuWutV4PXCyFWIQQQjiIo9rcuyqldimlliulmue3k1JqolIqTCkVduHCBQedWgghxLUckdz/BOporVsD/wN+zG9HrfVsrXUHrXWHypUrO+DUQggh8lLs5K61vqy1vmK7vQzwVEpVKnZkQgghrluxk7tSqppSStlud7IdM7q4xxVCCHH9Ch0to5T6GugDVFJKRQAvAp4AWutZwChgklIqFUgAxmqtdYlFLIQQolCFJnet9Z2FPP4+ZqikEEIIFyEzVIUQwoIkuQshhAVJchdCCAuS5C6EEBYkyV0IISxIkrsQQliQJHchhLAgSe5CCGFBktyFEMKCJLkLIYQFSXIXQggLkuQuhBAWJMldCCEsSJK7EEJYkCR3IYSwIEnuQghhQZLchRDCgiS5CyGEBUlyF0IIC5LkLoQQFiTJXQghLEiSuxBCWJAkdyGEsCBJ7kIIYUGS3IUQwoIkuQshhAVJchdCCAuS5C6EEBYkyV0IISxIkrsQQliQJHchhLAgSe5CCGFBktyFEMKCJLkLIYQFSXIXQggLkuQuhBAWJMldCCEsSJK7EEJYUKHJXSn1mVIqUim1J5/HlVLqPaXUEaXUbqVUO8eHKYQQoijsuXL/HBhUwOODgYa2n4nAzOKHJYQQojgKTe5a6/XAxQJ2GQF8qY0tQAWlVHVHBSiEEKLoHNHmHgKcynY/wrZNCCGEk5Rqh6pSaqJSKkwpFXbhwoXSPLUQQtxQHJHcTwO1st2vaduWi9Z6tta6g9a6Q+XKlR1waiGEEHlxRHL/CbjXNmqmCxCrtT7rgOMKIYS4Th6F7aCU+hroA1RSSkUALwKeAFrrWcAyYAhwBIgHxpdUsEIIIexTaHLXWt9ZyOMamOywiIQQQhSbzFAVQggLkuQuhBAWJMldCCEsSJK7EEJYkNOS+1+nY1m4PcJZpxdCCEtz6pX7G78ccObphRDCspya3C/EJWFGUgohhHAkp7e5/xF+ydkhCCGE5Tg9uV+8muTsEIQQwnKcntwTUtKcHYIQQliO85N7crqzQxBCCMtxenK/FJ/s7BCEEMJynJ7cE5KlWUYIIRzN6ck9KVWSuxBCOJoLJHdpcxdCCEdzenL/cvMJVu077+wwhBDCUpye3AEmfBnm7BCEEMJSXCK5CyGEcCyXSe5p6VJjRgghHMVlkvs7qw7ywuI9JEsHqxBCFJvTk/tjNzUA4IM1R/ly8wmW7Drj5IiEEKLsc2pyH9qqOiPbheTYNn3pPrTWRMYl0vLFFURdkcJiQghRVE5N7vd2qUO9yv45tl2KT6Hu08vo+9Y64pJSGT/nDydFJ4QQZZdTk7ubm8r3sStJqQDUDCpXWuEIIYRlODe5K5PcR3eome8+hyOvcNWW6DM4cvWmtQcj2Xos2mHHE0IIV+Dk5G5+twwpn7ntpeHNc+xzJPIKvf+9hh0nL9H3rbWs3HuOuk8vI3TqUofEcP+cPxgzewvf/nGS7ScuciQyziHHFUIIZ3Jqcne3ZfcLV7LK/t7XLZQ9Lw3ktrYhvHJrCwCiriTz8YZjHIu6ysS52zP33XQkqljnzz62/l+L/uL2mZvp9856Ll4t2TLEpy7GEzp1KUcvXCnR8wghblwu0Szzt74Ncmz39/bg3TFtuKdLHZ7s3wgAL/fcod79yVae/3EPY2dvZvuJi0U+f341bdq9sor/rj5c5OPZ69YPNgFw89vrSuwcQogbm1OTuy2345lH4s7QqlYFAH7cmff497lbTrDl2EVun7k5sxO2MHM3hxM6dSmPfGW+BdSr5Jdrn3dXHyIxJY29Z2JJTk0nPjmVd1aZbcXVu3HlzNuhU5eSkiYTt4QQjuUSzTIA3z/ajWVTeubap3HVgBz3w57rR/iMoQxtVT3Xvi1eXJGr8zUvzy/em+P+9492Y1Dzamz45005tnef8RtD39tIo+eWM3fzCd779TBzNoXnOt6uUzGZnbKvLt3HmgORhcaQXXGu4HediuGrLSeu+/lCCGvycObJM5plANrVDspzn6qB3pm3/zu2DZX8zf3XR7Zk6e6zALx5eyv+uWg3AO+sOsTzw5oVeF5Pd0VKmmlv/8fAxlTw9WLWPe0BqFfZj2MXrgIQna3t/fXlBwA4eTE+1/FG2JpZOtQJIuzEJT7ecJzwGUPzPf+lq8m0CAnE39uDLccucvJiPKlp6XgU8A0mL68t28/s9ccA8PJwY/W+88y+t0ORjiGEsCaXGC1TEKUUCx7uyo+TuzOiTdZs1kAfT964vSULH+nK6I61Mrd/uvE4oVOXEjp1KZFxibmOl30Y5e5pA5h8U872/pl3t6dDnbw/aAC+3nYyR9NMarYmlbATlzJvX0lKZdpPe/Ns19975jK+Xh58M7Erb9/RGoDJ8//M95zXGj1rM6FTl2YmdoB/LtzNyn3nZUavEAJwenK3I7sDnepWpI2t7T27MR1r0yG0IkCuJhWATq/+ypJdZzKLkSUkp1H36WWZV+2BPp65ntO4WgALJ3XLM44Ab/NFZ+3BC7z/22FCpy7lgS/yrkXf4sUVfP57eK5a9YkpaUTGJbHtuOkA7lzPxL9i73lmrj2a57EyYj8Xm0iH6avYFp5/53HGcYUQN7YykdztUauiL0dfG5Jr+9++3sG/Vxzgj/CLmU03AGM61Mq1b3ZHXxvCzU2qMH9CZ3a+0J9jrw3hj+f6oRR8/2cEb608BMD6QxcKje1yYkrm7flbTwJkjgKqGeTL329uCMAbvxwgPjmVuz7ewk+7zhAbn/W8pi/8QpfXfyXqSsHDNB+d9yeRl3N/YxFC3FiUI2d7FoV39Yb68J5d1A72dehxR3+0mW3HL9K7UWXWFZB4j78+BHUdHy75TZ5a81QfUtLSaVjFn7pPL8vx2FMDGvFY34Y5nr/v5YH4emV1eWRs71Y/mN+PZs2YzWi7z++8LwxrxtwtJzgedTXH9h3P9yfIz6sof5oQogxQSm3XWhfauebk2jKOP+a3E7sQPmMoH97dLt99/ny+/3UldoBJfepn3p5iG58/oFlV6lbyo1HVAJRSrHqiFwBvjmoFwFsrD5Gerhn2vw2Zz82e2AHm3N8RIEdiBzgSGZdnuYUm1QL47cne3N8tlDn3d+TlEc15eUTW7N7B/93A7oiY6/obhRBln2WaZTJkJG0/bw/CZwzl1yd7U728T+bjHeoEUbEYV7T/HNiYRlVNJctH+tTn+OtDco1QaVg1gPAZQxndoRYdQ03nbL1nlrHn9OV8j3tTkyp5bv/2j1PsiogF4NE+9dn70kD8vNx5vF8j6lX2x81NEVrJj3u7hjKuc9akr3OXExn+/ib+s/rQdf2dJ6Kv8uOO0wAkp6ZLWQYhyhinNsucOPAX1bIl3pKWmJJGSlo6AXl0pJaU1fvO81AeC4DnNVQyNj6F1i+vBODYa0Oo90zO5p2RbUN4Z0ybQs/Z9621HMvWTFPQsMy8PDw3jBV7c4/yGdGmBjWDyvFEv0ZFHrYphHAMhzbLKKUGKaUOKqWOKKWm5vH4/UqpC0qpnbafh+w5bkk0yxTEx9O9VBM7QL9mVZn3UOfM+61rVeA/+STo8r6eDG9dg/HdQ3FzU/h45nyBHu/XyK5zfvtwVxZN6sbQlmai1/pDFxj0n/VM/3lfoc/dfuJSnokdYPHOM3yw5igNnl1O6NSlHJPaOEK4rEKv3JVS7sAhoD8QAfwB3Km13pdtn/uBDlrrx+w9sXf1hvr04T2Zk5Ks7kJcEkG+nkW64s1+1b/6/3rRoEpAIc/IaePhKMZ9ujXHtsKu4otSbbNRVX9WPtG7SDEJIYrH3it3e2aodgKOaK2P2Q78DTACKPwysBAl0ebuqioHFP1DrF+zqnw+viNtawVR3rfo3zja5zEZ69D5OAa8ux4wk8ha16rAoObVuK9bKJcTsoZeBvp40KVeMCv3nefjeztwJSmFJ7732eTxAAATsElEQVTddc2xrpCUmoa3h3uRYxNClCx7rtxHAYO01g/Z7t8DdM5+lW67cn8duIC5yn9Ca32qoON6V2+oI4/uu66kJezX4JllpKYXrV/F39uDbyZ2oXG1AOKT0yhfzvwbxSenMn/rSaYv3Z+5b69GlflifMcCRx8lJKeRkp6e56SxgqSmpbP7dCzhUVe5rW3IdY9wEsJKSnso5BIgVGvdClgFfJFPUBOVUmFKqTAAJX1yJS7suX6MaFODx64ps1CQ2fe0p0VIeTzd3TITO5jhmw/1rMec+ztSwfahvP7QhVzj+q/1j4W76PTqal5eso8Ve89x9ydbaDltBX+evESDZ5bxy56zeT5v2P82MvLD3/m/Bbt4d9X1jfoR4kZlz5V7V2Ca1nqg7f7TAFrr1/PZ3x24qLUun9fjGbyrN9TR4fvx93Zq7bIbSkZ7esuQ8ky+qQENqvizIOwUyanpfP57OAAV/bzY9szNhfYNpKdrOr32a2Ytm01T+xJSIfd6t0mpaTR+7pcCj1W3kh9rnuqTb7wZijrqRwgrcmSb+x9AQ6VUXeA0MBa465qTVddaZ1x+DQf2Ywd3+Zpdqna+0J+txy8ysHm1zG3PDGkKwMDm1Yi6ksQtrWvYdSw3N0XYc/2Ys+k4Ly3ZR/cZv7Ht2Zup7O+do/nk2klZeTkedZU1ByJzjPXXWuPt4UZSalZhtugrSQTfIB3wQhRXoQ0jWutU4DFgBSZpL9Ba71VKvayUGm7bbYpSaq9SahcwBbjfnpNLbi9dFXy9ciT27LrWD7Y7sWc3tmPtzNudXv2VBz7/I8fjFy7nrlL5xu0t+fP5/nz1YGdm2mYSj//8Dxo+u4wlu8yiLCv3nScpNR1/bw/m24aStp++usjxCXGjcuokpisRBwtchUmUDacuxtPzzTWZ9zc/3Zfq5cuhtc5sj3/1thYkp6YzvnvdXM+/77Nt+dYBal2zPPMndKH5iysA2PbszWgNW49fZPh1fBgVR0apZx9PGR0knMfeZhmnJvf404dyrMYkyq6JX4ax0la7vlqgD53rVWRxtqURj7w6ON92/PjkVJq9sCLPxzIKrO05Hcuw/21kcItqLN9zDjBlnmtVdGzhOTALp/+44zTD29TIvPj4/UgUd31i5gwcfW2IvG+F05SNwmHy/8MyXrilGc8OaYqnu+Lc5cQcif2O9jUL7KD19TJ1gI6/PoRR7WvmegygeY1Aalf0zUzsQI5vC4605kAkT363i3eyjdDJSOwAs9blX3dfCFfh5AWyJbtbRc0gXyb0qsffbKWNM3SoE8S/Bjex6xhKKd66ozW7pw1g6zM3c/jVwTkeu7YcA8DpmITiBZ6HjFnBn2wwK12lXrOA+b9XHMxRa18IVyQN3sKh7uyU1cEaPmMoCyd1K3KJiUAfT6oG+uTqj3l+WDOaVg9k1RO9WPBwV8AsYn42NoGLV5PZnm2Zw2tFXMq99m1eYrPN0k1J03yw5ggX480CKU9n+5Bq/fJKzsY6/oNFCEdxWpu7T/WGOvHsYaecW5SstHRNutYl2lmevbM2u10vDiDQxyPzsRrlfXh+WDMmzfuTh3vX4+nBTfM95snoeHr9O/+mnpeGN6d/s6p0m/EbAO+Mbs3IdjXz3V+IklAm2tyFNbm7qRIfBaWU4vjruZdVfGvFQT7MthbtmdhEJs0zi49/tO4Y0VeS2Hosms228fcx8clE2yZirdib1Z6/aWpfWoQE5jh2n8aVqVGhHIdfHUw5T3d22+rsC+GKnDY9VNrbRXHl9R6au+VEgc954Iswdp0yK1Rte/ZmOr36KwAHXhmUY+HxGuV9+H5Sdxo9tzxzW51gPwA83d1oV6cCaw9GsvZgZXo1rIxbMUYHnItNxM+79MtRC2uTK3dRpv36pCk5PLRldWaMbJm5PdAnawTO0ik9eMJWCz8jsQOZiR2gyfO/sMo2lLN8OU+UUnh5uHFo+mDcFDzSO2t5RYDGVQMJj47n/jl/8NKSvYROXZrvWP1rO2Sz23g4ii6v/0rLaStJK2KBNyEK4rQ290bNW+tDe3cVvqMQdtJa8/GGY3SrX4kWIblLG42a+TthBXS6ZrBn8fTwqKv0eWttru27XhiQo9Lp9hOXuH3m73zxQCd6N6qM1jrz2AfOXWbQf7LW1ZU2fGEPl29zDywnX0GFYymlmNirfp6JHeDDcVmLps/PtjpWdj0bVrKryTC0kl/mQujZjZm9mZRsV+prD0YCZhZu6NSl1H16GXGJZkRO9sQOsOOkLGguHMdpV+4dOnTQYWG51xYVoiSdjU3gTEwi7esE8c22k3SuF8z2E5fYd+YyI9uFUKuib44yx4VJTEljx8kYYuKT2RkRw0frzNj48BlDmfBlWGZTT3bPD2vGfV3r8MhX21m9P5Kpg5uw8XAUkXGJLJ3SU0pyiAK5fPkBSe7CahJT0mjyvClvPPfBTtzz6bbMx9b/4yb6v7sus8plvcp+tKlVga3HLrJpal8+WneU15cfAOC7R7pyx6zNAEwd3CRXe7+4sbl8s4wQVuPj6Z5Zlz4jsd/fLZSvJ3ShdrAvB6cPZv4E0xx07MJVvv/zNDG2CVL9mlXNPE5GYgeYsfxAjmaeay3dfZbQqUsJnbqUpNQ0R/9JogyT5C6EA4UG5yxk9mif+nStH5x5v1v9SozrkjWL92qyScj1K/uz5LEeBPjkHp3c8NnlhE5dyuKdp3M9tizbKlYj3t9U7PiFdUhyF8KBlFLseL4/LUPK89zQplQJ9Mm1z/RbW2YOzZzQM6sEcsua5TOXQ2xfJ4jd0wbkWPPg79/s5I5ZvxMTn0xGc+raA5GZjx84F8fURbtL4s8SZZC0uQvhQlLS0nlrxUEm921AoI8nF68m0+6VVYU+75bWNTIXOvnfnW25pXUNElPScLON1xfW4chl9oQQpcTT3Y2nh2TVv6no58VvT/Ym2N+b5NR0Or6aezWquQ92onv9SlQN8OaTjcf529c78HRXPPXdbq4kpXL0tSEkpKQVab3i+ORUfDzcM2feJiSncTY2gXqV/Yv/R4pSIVfuQpQhyanpXIpP5vFvdtIxNIj/G9A4x+Oz1h1lhm3UzbVW/19vvv3jJO5ubjw5oBHJqen42RJ+errGzU2xYu85gny9GP3RZu7sVIvXR7YC4O2VB/nfb0eYfFN9/jHQvhLOomTIUEghblALt0fw1Hf2zf6eM74jU77eQVxiKq1rlmfXNcXQ3hzViibVApi17ijL/jKF1QK8PfjrpYEOj1vYR5K7EILUtHQ2H4vOMebeURZP7s7VpFS6Najk8GOL/Mk4dyEEHu5u9GxYmQOvDKJvkyo83q8hR14dTEiFcvk+58irgwmfMZS372idY3uLkEB+sxVqAxjxwSbu+mQrLy/Zx9srDxZa+OyjdUeZv/Vk5v24xJTMcf550VrLilfFIFfuQtzADp+PY1dELP2aVuGHHae5q3NtvD3cMx+PjU+hnJd7jhE3e8/E8uDnYZy7nJjreFP6NmB0x1r0eGNN5qidE9FXOXz+SubyhQAj24Xw/Z9m3P5H97SnSbWAzJLK6emazzYdZ9+Zy3y/4zT3dwvl89/DAVj5RC9CKpTL7Cu4EUmzjBCixJyJSeDlJfs4dD6OY1FX893v+OtDaPDscrvKGf/yeM9cxdQKM/+hznStH3xDrQ8hyV0IUeK01hyJvELDqgFsOHwhV9t+sJ8X0Vezml7+O7YNT3//F/HJaVT08+KL8Z245f2Ndp2re4NgNh2JzrW9Vc3yjOlYi7s717E77vR0zXOL9xBSoRwP9qhL5OUkal8zu9hVSXIXQpS641FXGffJVt4d04Zxn2wl2VYX59uJXdhy7CKTb6qPRx5VLxfvPM3fv9kJwKxx7RnUolqex09P10TGJbH+0AWiribx5i8Hczwe6OPB0ik9qVXRl/R0TXJaOj6e7sTGp7Ag7BT1Kvtxc9OqPP7NDn7ceSbHc+sE+3JT4yqZTUCHpg9mwpdhnL+cyPK/9yzw28GqfedpVbM8VW0zkrXWnLqYkO8HRlJqGi/8uJdvw05xaPpguyaaLfvrLKv2nec/Y9tKchdCOE/GguO3tQ3h3TFtSuQcvx+N4q6Pt5bIsbN78/ZW1KhQjhoVfPD18qCSv1fmh9TuiBiG2+r6PNC9Lnd3qc3Nb68DzDeXW1rXoG3tCoxoE0JY+EXWHIzkgzVHc53jv2PbMKJNSL4xhE5dCsCJN4ZJchdC3BiuJqXS/MUVRXrOnZ1q88+BjfnXot10qRfMyz/vA2B46xpU9PPKvIIvSEiFcpyOSbiekPO17dmbqRLgw6HzcXi5u+HhrujxxprMxyW5CyFuKGdjE/DxcCc2IYU1ByMJDfajU92KfLE5nA51KtKpbkXS0jUHzl0m4lICNzepkquJ6ExMAtXL+6CUYvj7G9kdEcuHd7fj0Xl/Fnr+fw1qwhu/ZM0OfmVEc55fvDfPfb083PhmYhfqVfKjzcu5awdlH02U3cDmVZl9b0dJ7kII4Qipael4uLtxJSmVlNR0jly4wu6IWE5EX+WhHvUy29bDo65SwdeTdG3qAl1OTOGnnWcY2S6EhdsjeGHxXv6aNoAAn9yrfSWlpjFnU3i+5SMy1vaVDlUhhCiDIuMSmfDldh7oHsqINiHsOR1LXGJq5roAUhVSCCHKoCoBPiye3D3zfn4LvhdGyg8IIYQFSXIXQggLkuQuhBAWJMldCCEsSJK7EEJYkCR3IYSwIEnuQghhQXYld6XUIKXUQaXUEaXU1Dwe91ZKfWt7fKtSKtTRgQohhLBfocldKeUOfAAMBpoBdyqlml2z24PAJa11A+Bd4A1HByqEEMJ+9ly5dwKOaK2Paa2TgW+AEdfsMwL4wnZ7IXCzupGWRhFCCBdjT3IPAU5lux9h25bnPlrrVCAWCHZEgEIIIYquVGvLKKUmAhNtd5OUUntK8/wOUgmIcnYQ16Esxl0WYwaJu7SVxbiLE7Nd6wnak9xPA7Wy3a9p25bXPhFKKQ+gPJBrsUOt9WxgNoBSKsyeymauRuIuPWUxZpC4S1tZjLs0YranWeYPoKFSqq5SygsYC/x0zT4/AffZbo8CftPOqiUshBCi8Ct3rXWqUuoxYAXgDnymtd6rlHoZCNNa/wR8CsxVSh0BLmI+AIQQQjiJXW3uWutlwLJrtr2Q7XYicEcRzz27iPu7Com79JTFmEHiLm1lMe4Sj9lpKzEJIYQoOVJ+QAghLMgpyb2wcgbOpJQKV0r9pZTaqZQKs22rqJRapZQ6bPsdZNuulFLv2f6O3UqpdqUY52dKqcjsw0mvJ06l1H22/Q8rpe7L61ylEPc0pdRp22u+Uyk1JNtjT9viPqiUGphte6m9h5RStZRSa5RS+5RSe5VSf7dtd+nXu4C4Xf319lFKbVNK7bLF/ZJte11beZMjypQ78bJtz7f8SX5/TynH/blS6ni217uNbXvJvk+01qX6g+mUPQrUA7yAXUCz0o6jgPjCgUrXbHsTmGq7PRV4w3Z7CLAcUEAXYGspxtkLaAfsud44gYrAMdvvINvtICfEPQ14Ko99m9neH95AXdv7xr2030NAdaCd7XYAcMgWm0u/3gXE7eqvtwL8bbc9ga2213EBMNa2fRYwyXb7UWCW7fZY4NuC/h4nxP05MCqP/Uv0feKMK3d7yhm4muzlFb4Abs22/UttbAEqKKWql0ZAWuv1mJFJxYlzILBKa31Ra30JWAUMckLc+RkBfKO1TtJaHweOYN4/pfoe0lqf1Vr/absdB+zHzMp26de7gLjz4yqvt9ZaX7Hd9bT9aKAvprwJ5H698yp/kt/fU9px56dE3yfOSO72lDNwJg2sVEptV2ZGLUBVrfVZ2+1zQFXbbVf7W4oapyvF/5jtq+lnGc0buGDctq/8bTFXZWXm9b4mbnDx11sp5a6U2glEYpLbUSBGm/Im18aQX/kTp8ettc54vV+1vd7vKqW8r437mvgcErd0qObWQ2vdDlMFc7JSqlf2B7X53uTyQ4zKSpw2M4H6QBvgLPC2c8PJm1LKH1gEPK61vpz9MVd+vfOI2+Vfb611mta6DWZGfCegiZNDssu1cSulWgBPY+LviGlq+VdpxOKM5G5POQOn0Vqftv2OBH7AvLHOZzS32H5H2nZ3tb+lqHG6RPxa6/O2/xTpwMdkfXV2mbiVUp6YBDlPa/29bbPLv955xV0WXu8MWusYYA3QFdNskTE3J3sMmfGpnOVPXCHuQbbmMa21TgLmUEqvtzOSuz3lDJxCKeWnlArIuA0MAPaQs7zCfcBi2+2fgHttvd5dgNhsX9OdoahxrgAGKKWCbF/NB9i2lapr+iluw7zmYOIeaxsNURdoCGyjlN9DtvbbT4H9Wut3sj3k0q93fnGXgde7slKqgu12OaA/pr9gDaa8CeR+vfMqf5Lf31OacR/IdgGgMP0E2V/vknufFLUH1hE/mF7iQ5h2tGedEUM+cdXD9K7vAvZmxIZpv/sVOAysBirqrN7xD2x/x19Ah1KM9WvMV+oUTJvcg9cTJ/AApqPpCDDeSXHPtcW12/aGr55t/2dtcR8EBjvjPQT0wDS57AZ22n6GuPrrXUDcrv56twJ22OLbA7xg214Pk5yPAN8B3rbtPrb7R2yP1yvs7ynluH+zvd57gK/IGlFTou8TmaEqhBAWJB2qQghhQZLchRDCgiS5CyGEBUlyF0IIC5LkLoQQFiTJXQghLEiSuxBCWJAkdyGEsKD/B/G6sU23pLxcAAAAAElFTkSuQmCC\n", "text/plain": [ "

" ] }, "metadata": {}, "output_type": "display_data" }, { "name": "stdout", "output_type": "stream", "text": [ "Better model found at epoch 0 with accuracy value: 0.33958810567855835.\n", "Better model found at epoch 4 with accuracy value: 0.35972538590431213.\n", "Better model found at epoch 5 with accuracy value: 0.38627001643180847.\n", "Better model found at epoch 8 with accuracy value: 0.40869563817977905.\n", "Better model found at epoch 9 with accuracy value: 0.4453089237213135.\n", "Better model found at epoch 11 with accuracy value: 0.4526315927505493.\n", "Better model found at epoch 12 with accuracy value: 0.45537757873535156.\n", "Better model found at epoch 13 with accuracy value: 0.4864988625049591.\n" ] } ], "source": [ "learn.fit_one_cycle(epochs_all, max_lr = slice(1e-5, 1e-4), wd=0.05, \n", " callbacks=[SaveModelCallback(learn, monitor='accuracy',\n", " mode='max', name='Best_model_sz224_stage2_resnet50_fold5'),\n", " ReduceLROnPlateauCallback(learn, monitor='accuracy', mode='max', patience = patience)])\n", "\n" ] }, { "cell_type": "code", "execution_count": 74, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "2185\n" ] }, { "data": { "text/plain": [ "True" ] }, "execution_count": 74, "metadata": {}, "output_type": "execute_result" } ], "source": [ "learn.load('Best_model_sz224_stage2_resnet50_fold5')\n", "\n", "interp5 = ClassificationInterpretation.from_learner(learn)\n", "print(interp5.confusion_matrix().sum())\n", "\n", "losses,idxs = interp5.top_losses()\n", "len(data.valid_ds)==len(losses)==len(idxs)" ] }, { "cell_type": "code", "execution_count": 75, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.6755170098556844" ] }, "execution_count": 75, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix5 = interp5.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix5.sum(axis=0) - np.diag(confusion_matrix5) \n", "FN = confusion_matrix5.sum(axis=1) - np.diag(confusion_matrix5)\n", "TP = np.diag(confusion_matrix5)\n", "TN = confusion_matrix5.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "\n", "TPR.mean()" ] }, { "cell_type": "code", "execution_count": 76, "metadata": {}, "outputs": [ { "data": { "text/html": [ "Total time: 00:51

" ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "0.6691429111511357" ] }, "execution_count": 76, "metadata": {}, "output_type": "execute_result" } ], "source": [ "interp5_tta = ClassificationInterpretation.from_learner(learn, tta=True)\n", "\n", "confusion_matrix5_tta = interp5_tta.confusion_matrix()\n", "\n", "\n", "FP = confusion_matrix5_tta.sum(axis=0) - np.diag(confusion_matrix5_tta) \n", "FN = confusion_matrix5_tta.sum(axis=1) - np.diag(confusion_matrix5_tta)\n", "TP = np.diag(confusion_matrix5_tta)\n", "TN = confusion_matrix5_tta.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "\n", "\n", "TPR.mean()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Analysis of the Cross Validated Results" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "We obtain a 66.8% balanced multiclass accuracy without augmentation and a 67.9% balanced multiclass accuracy using test time augmentation. Details of the other classification metrics are presented below." ] }, { "cell_type": "code", "execution_count": 77, "metadata": {}, "outputs": [], "source": [ "confusion_matrix = confusion_matrix1 + confusion_matrix2 + confusion_matrix3 + confusion_matrix4 + confusion_matrix5" ] }, { "cell_type": "code", "execution_count": 78, "metadata": {}, "outputs": [], "source": [ "confusion_matrix_TTA = confusion_matrix1_tta + confusion_matrix2_tta + confusion_matrix3_tta + confusion_matrix4_tta + confusion_matrix5_tta" ] }, { "cell_type": "code", "execution_count": 79, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "array([[ 206, 29, 53, 9, 22, 8, 0],\n", " [ 37, 372, 37, 17, 22, 23, 6],\n", " [ 60, 24, 666, 2, 196, 147, 4],\n", " [ 12, 9, 9, 65, 11, 9, 0],\n", " [ 25, 14, 127, 16, 657, 266, 8],\n", " [ 35, 106, 445, 79, 1136, 4863, 41],\n", " [ 3, 4, 1, 0, 8, 7, 119]])" ] }, "execution_count": 79, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix" ] }, { "cell_type": "code", "execution_count": 80, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "array([[ 229, 25, 40, 8, 17, 8, 0],\n", " [ 41, 388, 33, 12, 11, 20, 9],\n", " [ 77, 21, 669, 2, 167, 162, 1],\n", " [ 11, 10, 7, 65, 11, 11, 0],\n", " [ 38, 9, 131, 14, 616, 295, 10],\n", " [ 45, 98, 432, 37, 968, 5082, 43],\n", " [ 2, 6, 1, 0, 10, 7, 116]])" ] }, "execution_count": 80, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix_TTA" ] }, { "cell_type": "code", "execution_count": 81, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "10015" ] }, "execution_count": 81, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix.sum()" ] }, { "cell_type": "code", "execution_count": 82, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "10015" ] }, "execution_count": 82, "metadata": {}, "output_type": "execute_result" } ], "source": [ "confusion_matrix_TTA.sum()" ] }, { "cell_type": "code", "execution_count": 83, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "True Positive Rate [0.629969 0.723735 0.606005 0.565217 0.590296 0.72528 0.838028]\n", "True Negative Rate [0.982246 0.980423 0.92463 0.987576 0.843294 0.861027 0.994024]\n", "Positive Predictive Value [0.544974 0.666667 0.497758 0.345745 0.320175 0.913583 0.668539]\n", "Negative Predictive Value [0.987444 0.984985 0.950098 0.994912 0.942735 0.607417 0.997662]\n", "False Positive Rate [0.017754 0.019577 0.07537 0.012424 0.156706 0.138973 0.005976]\n", "False Negative Rate [0.370031 0.276265 0.393995 0.434783 0.409704 0.27472 0.161972]\n", "False Discovery Rate [0.455026 0.333333 0.502242 0.654255 0.679825 0.086417 0.331461]\n", "Accuracy [0.970744 0.967249 0.889666 0.982726 0.815177 0.770145 0.991812]\n" ] } ], "source": [ "\n", "#https://stackoverflow.com/questions/31324218/scikit-learn-how-to-obtain-true-positive-true-negative-false-positive-and-fal\n", "\n", "FP = confusion_matrix.sum(axis=0) - np.diag(confusion_matrix) \n", "FN = confusion_matrix.sum(axis=1) - np.diag(confusion_matrix)\n", "TP = np.diag(confusion_matrix)\n", "TN = confusion_matrix.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "# Specificity or true negative rate\n", "TNR = TN/(TN+FP) \n", "# Precision or positive predictive value\n", "PPV = TP/(TP+FP)\n", "# Negative predictive value\n", "NPV = TN/(TN+FN)\n", "# Fall out or false positive rate\n", "FPR = FP/(FP+TN)\n", "# False negative rate\n", "FNR = FN/(TP+FN)\n", "# False discovery rate\n", "FDR = FP/(TP+FP)\n", "\n", "# Overall accuracy\n", "ACC = (TP+TN)/(TP+FP+FN+TN)\n", "\n", "\n", "print(\"True Positive Rate \", TPR)\n", "print(\"True Negative Rate \", TNR)\n", "print(\"Positive Predictive Value \", PPV)\n", "print(\"Negative Predictive Value \", NPV)\n", "print(\"False Positive Rate \", FPR)\n", "print(\"False Negative Rate \", FNR)\n", "print(\"False Discovery Rate \", FDR)\n", "print(\"Accuracy \", ACC)" ] }, { "cell_type": "code", "execution_count": 84, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.6683617121946673" ] }, "execution_count": 84, "metadata": {}, "output_type": "execute_result" } ], "source": [ "TPR.mean() #without augmentation" ] }, { "cell_type": "code", "execution_count": 85, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.9390313951117332" ] }, "execution_count": 85, "metadata": {}, "output_type": "execute_result" } ], "source": [ "TNR.mean()" ] }, { "cell_type": "code", "execution_count": 86, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "True Positive Rate [0.700306 0.754864 0.608735 0.565217 0.553459 0.757942 0.816901]\n", "True Negative Rate [0.977911 0.982212 0.92777 0.992626 0.866996 0.848036 0.993619]\n", "Positive Predictive Value [0.51693 0.696589 0.50952 0.471014 0.342222 0.909937 0.648045]\n", "Negative Predictive Value [0.989762 0.986678 0.950586 0.994938 0.939501 0.633634 0.997357]\n", "False Positive Rate [0.022089 0.017788 0.07223 0.007374 0.133004 0.151964 0.006381]\n", "False Negative Rate [0.299694 0.245136 0.391265 0.434783 0.446541 0.242058 0.183099]\n", "False Discovery Rate [0.48307 0.303411 0.49048 0.528986 0.657778 0.090063 0.351955]\n", "Accuracy [0.968847 0.970544 0.892761 0.987718 0.832152 0.787718 0.991113]\n" ] } ], "source": [ "FP = confusion_matrix_TTA.sum(axis=0) - np.diag(confusion_matrix_TTA) \n", "FN = confusion_matrix_TTA.sum(axis=1) - np.diag(confusion_matrix_TTA)\n", "TP = np.diag(confusion_matrix_TTA)\n", "TN = confusion_matrix_TTA.sum() - (FP + FN + TP)\n", "\n", "# Sensitivity, hit rate, recall, or true positive rate\n", "TPR = TP/(TP+FN)\n", "# Specificity or true negative rate\n", "TNR = TN/(TN+FP) \n", "# Precision or positive predictive value\n", "PPV = TP/(TP+FP)\n", "# Negative predictive value\n", "NPV = TN/(TN+FN)\n", "# Fall out or false positive rate\n", "FPR = FP/(FP+TN)\n", "# False negative rate\n", "FNR = FN/(TP+FN)\n", "# False discovery rate\n", "FDR = FP/(TP+FP)\n", "\n", "# Overall accuracy\n", "ACC = (TP+TN)/(TP+FP+FN+TN)\n", "\n", "\n", "print(\"True Positive Rate \", TPR)\n", "print(\"True Negative Rate \", TNR)\n", "print(\"Positive Predictive Value \", PPV)\n", "print(\"Negative Predictive Value \", NPV)\n", "print(\"False Positive Rate \", FPR)\n", "print(\"False Negative Rate \", FNR)\n", "print(\"False Discovery Rate \", FDR)\n", "print(\"Accuracy \", ACC)" ] }, { "cell_type": "code", "execution_count": 87, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.67963208445167" ] }, "execution_count": 87, "metadata": {}, "output_type": "execute_result" } ], "source": [ "TPR.mean() # with test time augmentation" ] }, { "cell_type": "code", "execution_count": 88, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.9413101678032485" ] }, "execution_count": 88, "metadata": {}, "output_type": "execute_result" } ], "source": [ "TNR.mean()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Limitations\n", "\n", "\n", "The test data from the ISIC 2018 is not available. Althought we are confident that our cross-validation strategy is reliable, we are not able to submit the results to the competition and get a leaderboard standing.\n", "\n", "We only have one GPU NVIDIA 2080 with memory size of 8GB. The use of multiple GPUs can speed the training process and increase accuracy by processing a larger batch size.\n", "\n", "We have made all efforts to make the training process as deterministic as possible using the seed_everything function. However, training of CNN still carries a stochastic component, and the results may be replicated but not identically. \n", "\n", "There is great variability in the results for each one of the folds, with validation scores ranging from 62% to 79%. This may suggest that the weighted cross-entropy metric that we optimized may not be very stable, and other optimization metrics that focus on the detection of rare events, such as focal loss may be more beneficial for the training of the neural network." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Conclussions\n", "\n", "Computer vision algorithms have enormous potential in healthcare applications. This jupyter notebook illustrates the process of training a CNN model to model a dermatological dataset of seven skin conditions. Several other applications of this methodology have the potential to improve diagnosis accuracy and reduce healthcare costs." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## References\n", "\n", "[1] Tschandl, P., Rosendahl, C. & Kittler, H. The HAM10000 dataset, a large collection of multi-source \n", "dermatoscopic images of common pigmented skin lesions. Sci. Data 5, 180161 (2018). \n", "doi: 10.1038/sdata.2018.161 \n", "\n", "[2] He, Kaiming et al. Deep Residual Learning for Image Recognition. Microsoft Research. https://arxiv.org/abs/1512.03385v1\n", "\n", "[3] ISIC 2018 challenge Task 3: Lesion Diagnosis https://challenge2018.isic-archive.com/leaderboards/\n", "\n", "[4] Shang, Kailan. Applying Image Recognition to Insurance. Society of Actuaries. June 2018. https://www.soa.org/globalassets/assets/Files/resources/research-report/2018/applying-image-recognition.pdf\n", "\n", "[5] Pranav Rajpurkar, et al. CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning. https://arxiv.org/abs/1711.05225\n", "\n", "[6] Levy, Daniel, Jain, Arzav. Breast Mass Classification from Mammograms using Deep Convolutional Neural Networks. 2016. https://arxiv.org/abs/1612.00542\n", "\n", "[7] Esteva, Andre, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature volume 542, pages 115–118 (02 February 2017).\n", "\n", "[8] Howard, Jeremy and others, fastai, 2018, publisher= GitHub, https://github.com/fastai/fastai\n", "\n" ] } ], "metadata": { "kernelspec": { "display_name": "fastAI custom", "language": "python", "name": "fastai" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.0" }, "latex_envs": { "LaTeX_envs_menu_present": true, "autoclose": false, "autocomplete": true, "bibliofile": "biblio.bib", "cite_by": "number", "current_citInitial": 1, "eqLabelWithNumbers": true, "eqNumInitial": 1, "hotkeys": { "equation": "Ctrl-E", "itemize": "Ctrl-I" }, "labels_anchors": false, "latex_user_defs": false, "report_style_numbering": false, "user_envs_cfg": false }, "toc": { "base_numbering": 1, "nav_menu": {}, "number_sections": false, "sideBar": true, "skip_h1_title": false, "title_cell": "Table of Contents", "title_sidebar": "Contents", "toc_cell": false, "toc_position": { "height": "calc(100% - 180px)", "left": "10px", "top": "150px", "width": "291.883px" }, "toc_section_display": true, "toc_window_display": true }, "varInspector": { "cols": { "lenName": 16, "lenType": 16, "lenVar": 40 }, "kernels_config": { "python": { "delete_cmd_postfix": "", "delete_cmd_prefix": "del ", "library": "var_list.py", "varRefreshCmd": "print(var_dic_list())" }, "r": { "delete_cmd_postfix": ") ", "delete_cmd_prefix": "rm(", "library": "var_list.r", "varRefreshCmd": "cat(var_dic_list()) " } }, "types_to_exclude": [ "module", "function", "builtin_function_or_method", "instance", "_Feature" ], "window_display": false } }, "nbformat": 4, "nbformat_minor": 2 }