# Change-Aware Sampling and Contrastive Learning for Satellite Images This is the implementation for [CVPR 2023](https://cvpr2023.thecvf.com/) paper [Change-Aware Sampling and Contrastive Learning for Satellite Images](https://research.cs.cornell.edu/caco). Authors: [Utkarsh Mall](https://www.cs.cornell.edu/~utkarshm/), [Bharath Hariharan](http://home.bharathh.info/), [Kavita Bala](https://www.cs.cornell.edu/~kb/) ![CACO](teaser.png) ## Installation This codebase is built with and tested with python3.9, and torch 1.7.1. We highly recommend creating environment using anaconda. > conda create --name cacoenv python=3.9.13 -y After activating the environment: > conda activate cacoenv Install the following packages: >conda install cudatoolkit=11.0 -y >pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 -f https://download.pytorch.org/whl/torch_stable.html >pip install matplotlib pytorch-lightning==1.1.8 pytorch-lightning-bolts==0.3.0 scikit-learn rasterio lmdb pandas >pip installjupyter progressbar opencv-python Optionally also install >pip install wandb gym If running into error due to some library mismatch, please refer to `requirements.txt` for the correct versions. ## Quick inference Use the notebook for quick inference of on our pretrained model. > cd notebooks > python -m jupyter notebook --ip 0.0.0.0 --no-browser The notebook also provides instructions on how to download pretrained models. The pretrained models can be found at [this link](https://research.cs.cornell.edu/caco/checkpoints/) ## Training In order to retrain the model, we need to download the CACo data and then run training. A sample training procedure is provided in `sample_training.sh` > bash sample_training.sh The data can be found at [this link](https://research.cs.cornell.edu/caco/data/caco/) Refer to the `sample_training.sh` for downloading and placing data. To train a `resnet-18` model with `10k data` and `caco` loss run the following > python3 main_pretrain.py --data_dir ../data/clean_10k_geography/ --base_encoder resnet18 --batch_size 256 --data_mode caco --max_epochs 1000 --schedule 600 800 -d gereric_description ## A few more things will be released soon. - [ ] Code to download custom dataset using CACo sampling. - [ ] Setup for evaluating on other datasets, BigEarthNet, DynmamicEarthNet, OSCD, and Change Events. - [x] [1m training set](https://zenodo.org/records/10913174). ## Acknowledgements This repository follows the stucture from [SeCo](https://github.com/ServiceNow/seasonal-contrast) and uses same versions of libraries used.