Execution Instructions
-
Change current directory to: 'Code/' (where code files are located)
-
Prepare data in the given format and place it in : 'Data/' directory
-
Execution 3.1 python prepare_data.py It prepares and dumps the dataframes in a format which can be used by models and dataloaders to train and evaluate.
3.2 python train_tgif.py Trains the proposed tgif model(27) on the prepared data and dumps the best models in 'models/27/'
3.3 python eval_tgif.py Loads the best model and calculates weighted f1, recall and precision scores for both validation and test set. And saves the grouth truth and prediction csv in 'pred_csv/'