dingfengshi / react Goto Github PK

Hello, thank you for your excellent work. I have a question to ask you:
The video features are extracted by the feature extractor. Before the transformer encoder, it passes through an FC layer. What are the codes for feature processing and input to the encoder before the encoder? And what are the codes that the encoder outputs after the feature processing?

Thank you.

Hello! What is baseline In Table 3? Deformable detr(no two-stage)+roialign or Deformable detr(no two-stage)?

Hello, the installation part of the code really do not understand, can you be more detailed?

In the React/thumos_dataset.py, and in the getitem () function, if it is training mode now, and when the self.provide_contrastive_data is True, we will get sample_gt as well as pos_sample_segment, neg_sample_segment and candidated_segments.

When we get pos_sample_segment, neg_sample_segment and candidated_segments, as the code shows:
pos_sample_segment = pos_sample_segment * (len(pos_feat) / self.clip_len)
neg_sample_segment = neg_sample_segment * (len(neg_feat) / self.clip_len)
candidated_segments = candidated_segments * (gt_len / self.clip_len)
Their start and end points are scaled to a value relative to 256.
However, the start and end of sample_gt are still relative to its own snippet_num.

Should a similar scaling operation be done for sample_gt here?

where is the backbone which is the ReAct network?

I found the bsn_train_proposal_list.txt under React/thumos_dataset_meta only has 16 videos, and even the #15 misses some pr_boxes.

Could you please update the file as soon as possible?