{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "JwEAhQVzkAwA"
   },
   "source": [
    "# Convert a TensorFlow Model to OpenVINO™\n",
    "\n",
    "This short tutorial shows how to convert a TensorFlow [MobileNetV3](https://docs.openvino.ai/latest/omz_models_model_mobilenet_v3_small_1_0_224_tf.html) image classification model to OpenVINO [Intermediate Representation](https://docs.openvino.ai/latest/openvino_docs_MO_DG_IR_and_opsets.html) (OpenVINO IR) format, using [Model Optimizer](https://docs.openvino.ai/latest/openvino_docs_MO_DG_Deep_Learning_Model_Optimizer_DevGuide.html). After creating the OpenVINO IR, load the model in [OpenVINO Runtime](https://docs.openvino.ai/latest/openvino_docs_IE_DG_Deep_Learning_Inference_Engine_DevGuide.html) and do inference with a sample image.  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "QB4Yo-rGGLmV"
   },
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2ynWRum4iiTz"
   },
   "outputs": [],
   "source": [
    "import time\n",
    "from pathlib import Path\n",
    "\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from IPython.display import Markdown\n",
    "from openvino.runtime import Core"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Settings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The paths of the source and converted models.\n",
    "model_dir = Path(\"model\")\n",
    "model_dir.mkdir(exist_ok=True)\n",
    "\n",
    "model_path = Path(\"model/v3-small_224_1.0_float\")\n",
    "\n",
    "ir_path = Path(\"model/v3-small_224_1.0_float.xml\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Download model\n",
    "\n",
    "Load model using [tf.keras.applications api](https://www.tensorflow.org/api_docs/python/tf/keras/applications/MobileNetV3Small) and save it to the disk."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.applications.MobileNetV3Small()\n",
    "model.save(model_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6JSoEIk60uxV"
   },
   "source": [
    "## Convert a Model to OpenVINO IR Format\n",
    "\n",
    "### Convert a TensorFlow Model to OpenVINO IR Format\n",
    "\n",
    "Use Model Optimizer to convert a TensorFlow model to OpenVINO IR with `FP16` precision. The models are saved to the current directory. Add mean values to the model and scale the output with the standard deviation with `--scale_values`. With these options, it is not necessary to normalize input data before propagating it through the network. The original model expects input images in `RGB` format. The converted model also expects images in `RGB` format. If you want the converted model to work with `BGR` images, use the `--reverse-input-channels` option. For more information about Model Optimizer, including a description of the command-line options, see the [Model Optimizer Developer Guide](https://docs.openvino.ai/latest/openvino_docs_MO_DG_Deep_Learning_Model_Optimizer_DevGuide.html). For information about the model, including input shape, expected color order and mean values, refer to the [model documentation](https://docs.openvino.ai/latest/omz_models_model_mobilenet_v3_small_1_0_224_tf.html).\n",
    "\n",
    "First construct the command for Model Optimizer, and then execute this command in the notebook by prepending the command with an `!`. There may be some errors or warnings in the output. When model optimization is successful, the last lines of the output will include `[ SUCCESS ] Generated IR version 11 model.`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Construct the command for Model Optimizer.\n",
    "mo_command = f\"\"\"mo\n",
    "                 --saved_model_dir \"{model_path}\" \n",
    "                 --input_shape \"[1,224,224,3]\" \n",
    "                 --mean_values=\"[127.5,127.5,127.5]\"\n",
    "                 --scale_values=\"[127.5]\" \n",
    "                 --data_type FP16\n",
    "                 --model_name \"{model_path.name}\"\n",
    "                 --output_dir \"{model_path.parent}\"\n",
    "                 \"\"\"\n",
    "mo_command = \" \".join(mo_command.split())\n",
    "print(\"Model Optimizer command to convert TensorFlow to OpenVINO:\")\n",
    "display(Markdown(f\"`{mo_command}`\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run Model Optimizer if the IR model file does not exist\n",
    "if not ir_path.exists():\n",
    "    print(\"Exporting TensorFlow model to IR... This may take a few minutes.\")\n",
    "    ! $mo_command\n",
    "else:\n",
    "    print(f\"IR model {ir_path} already exists.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Test Inference on the Converted Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ie = Core()\n",
    "model = ie.read_model(ir_path)\n",
    "compiled_model = ie.compile_model(model=model, device_name=\"CPU\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Get Model Information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_key = compiled_model.input(0)\n",
    "output_key = compiled_model.output(0)\n",
    "network_input_shape = input_key.shape "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load an Image\n",
    "\n",
    "Load an image, resize it, and convert it to the input shape of the network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The MobileNet network expects images in RGB format.\n",
    "image = cv2.cvtColor(cv2.imread(filename=\"data/coco.jpg\"), code=cv2.COLOR_BGR2RGB)\n",
    "\n",
    "# Resize the image to the network input shape.\n",
    "resized_image = cv2.resize(src=image, dsize=(224, 224))\n",
    "\n",
    "# Transpose the image to the network input shape.\n",
    "input_image = np.expand_dims(resized_image, 0)\n",
    "\n",
    "plt.imshow(image);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Do Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "result = compiled_model(input_image)[output_key]\n",
    "\n",
    "result_index = np.argmax(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Convert the inference result to a class name.\n",
    "imagenet_classes = open(\"utils/imagenet_2012.txt\").read().splitlines()\n",
    "\n",
    "# The model description states that for this model, class 0 is background.\n",
    "# Therefore, add background at the beginning of imagenet_classes\n",
    "imagenet_classes = ['background'] + imagenet_classes\n",
    "\n",
    "imagenet_classes[result_index]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Timing\n",
    "\n",
    "Measure the time it takes to do inference on thousand images. This gives an indication of performance. For more accurate benchmarking, use the [Benchmark Tool](https://docs.openvino.ai/latest/openvino_inference_engine_tools_benchmark_tool_README.html) in OpenVINO. Note that many optimizations are possible to improve the performance. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_images = 1000\n",
    "\n",
    "start = time.perf_counter()\n",
    "\n",
    "for _ in range(num_images):\n",
    "    compiled_model([input_image])\n",
    "\n",
    "end = time.perf_counter()\n",
    "time_ir = end - start\n",
    "\n",
    "print(\n",
    "    f\"IR model in OpenVINO Runtime/CPU: {time_ir/num_images:.4f} \"\n",
    "    f\"seconds per image, FPS: {num_images/time_ir:.2f}\"\n",
    ")"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "OpenVINO 2021.3 PIP installer - PyTorch Image Segmentation.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}