The number of pairs does not match.

[xwjabc]

When I tried to extract the pairs from the training set (use annotations of non-zero style and create pairs using same pair_id & style but different source), it will have 394,661 pairs. However, the link indicates that there are 337,293 pairs, which is different from the number I have got. I wonder if there are some issues in my pair generation procedure.

[geyuying]

Is 394,661 refers to the number of image pairs or instance pairs?

[xwjabc]

Instance pairs. So the number in the link indicates the image pairs?

[geyuying]

Yes. I think your pair generation procedure is correct.

[xwjabc]

Thx! Now I can get 337,293 image pairs, which contain the user-shop image pairs and those images have at least one instance with style != 0. By the way, I wonder if in match r-cnn, the data loader directly samples pairs from the 337,293 image pairs?

[LeCongThuong]

Hi @xwjabc In your process of creating user-shop image pairs and those images have at least one instance with style != 0, did you meet the situation where instances with style !=0 exists only one?