DS4CG TNC

Installation

Some python libraries are required to use our pipeline. All the necessary python libraries are specified in requirements.txt. Run the following commend to install the libraries at once.

bash scripts/setup.sh

Datasets

iWildCam2018

As a pretraining step of our model, we first train the model on iWildCam2018 dataset. For detailed information about iWildCam2018, please refer to iWildCam2018 github repo. Run the following commend to download iWildCam2018 train, validation and test sets and corresponding annotation file. Note that download_path flag indicates the directory to download the dataset. If the flag is not provided, the dataset will be downloaded to $repo/data/wildcam directory. People who are using Gypsum do not need to download the dataset.

bash scripts/download.sh --download_path data/wildcam

TNC

TBD

Training

Outside of Gypsum:

bash scripts/run_train.sh --config resnet18 --tag test

In Gypsum:

sbatch -p titanx-short --gres=gpu:1 --output=out/test.out scripts/run_train.sh --config resnet18 --tag test

Evaluation

Outside of Gypsum:

bash scripts/run_eval.sh --config resnet18 --tag test

In Gypsum:

sbatch -p titanx-short --gres=gpu:1 --output=out/test.out scripts/run_eval.sh --config resnet18 --tag test