Prototyping repository for our solution to the AI-HERO Health Challenge Hackathon.
Run following commands from the root dir of this repository (tested with python=3.8):
pip install -r requirements.txt
pip install .
Inference can be run with the following command:
sbatch eval_conda.sh --save_dir [dir to save results] --csv_filepath [Filepath of csv] --data_dir [directory containing imgs of csv file] --ckpt_path [path to checkpoint]
Details:
- save_dir: Directory where predictions will be saved to as 'predictions.csv'
- data_dir: Directory to the folder containing the images of the csv file.
- csv_filepath: Path to the csv file containing the "images" in the first column and the first row is the header
- ckpt_path: Path to the checkpoint file loaded from Zenodo.