In the paper, we have proposed two different approaches;
- Single-task model
- Multi-task model
Run following commands to train, validate and evaluate multitask model on the challenge dataset.
- Model training:
python main.py --resume_model path/to/saved/model-weights - Model evaluation:
python main.py --test_model --resume_model path/to/saved/model-weights
The prediction file on publictest.csv will be generated for the post-evaluation phase of the competition.