Abstract: In deep metric learning, the training procedure relies on sampling informative tuples. However, as the training procedure progresses, it becomes nearly impossible to sample relevant hard ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback