Some question about the code.

[sqiangcao99]

Hi, thanks for the code. I have some questions regarding the code in tal_metrics.py, which is used to evaluate TAL performance. Specifically, I would like clarification on the ground truth files used for evaluation: thumos_from_oad_multiclasses.pkl and thumos14.json. What are the differences between these two files?

In the code, the GT file is loaded as follows:

if self.oad_gt_file is None:
    # thumos14.json
    self.ground_truth = load_gt_seg_from_json(ant_file, split=self.split, label=label, label_offset=label_offset)
else:
    # thumos_from_oad_multiclasses.pkl
    print('load OAD gt file for evaluation: ', self.oad_gt_file)
    self.ground_truth = mmcv.load(self.oad_gt_file)

Could you please provide an explanation of the distinctions between thumos14.json and thumos_from_oad_multiclasses.pkl?