Closed ousinkou closed 4 years ago
The train set, val set and test set are all cropped with the known bounding box. Did it mean an object algorithm should be added before attribute prediction in the real situation?
No. The bounding box can improve the training results but it is not necessary.
The train set, val set and test set are all cropped with the known bounding box. Did it mean an object algorithm should be added before attribute prediction in the real situation?