search

ikuinen / CMIN_moment_retrieval

Cross-Modal Interaction Networks for Query-Based Moment Retrieval in Videos
86 stars 20 forks source link

Bug in data loader? #14

Open ChenyunWu opened 3 years ago

ChenyunWu commented 3 years ago

In CMIN_moment_retrieval/dataloaders/clip_loader.py line 66 if label[1] == video.shape[0] - 1: Shouldn't it be: if label[1] == ori_video_len:? Since you changed the video and its length is now always self.max_num_frames

onlyonewater commented 3 years ago

I think the logic of this code feels strange!!!

ikuinen commented 3 years ago

In CMIN_moment_retrieval/dataloaders/clip_loader.py line 66 if label[1] == video.shape[0] - 1: Shouldn't it be: if label[1] == ori_video_len:? Since you changed the video and its length is now always self.max_num_frames

Thanks for reminding. We have corrected this in the subsequent use but have not updated this code. Even with a low probability, this may cause an error when length > self.max_num_frames and label[1] == self.max_num_frames - 1.

Starboy-at-earth commented 3 years ago

Dear ChenyunWu:

Could you please tell me how to process the splitted C3D features of Activity Caption dataset? It is weired five parts and I cannot figure out how to restore such files with suffixs of .part-XX.