/ORE

Primary LanguagePython

Recent Changes

  • Added --full_folder param to eval_linear

    • Take the entire "linear_prob_dataset" and does a 80/20 split into train and val in set_loaders line 90

  • Added label_to_supcat and set_meters

  • Added best_acc map in main for INat, be sure to change if needed for ORE

  • Changes to _average_binary_score to handle general lable_to_groups

  • Commented out high_resolution and low_resolutions pic handling in train method, besure sure to add that in and handle properly. Such as at line 486:

            # low_resolution_output = model(low_resolution_batch)
            # high_resolution_output = model(high_resolution_batch)
            # output = torch.cat((high_resolution_output, low_resolution_output), dim=0)
            output = model(image_list)