Accuracy is always zero

[LXYTSOS]

The accuracy is always zero, any idea of what might went wrong?

[bnu-wangxun]

The accuracy during training makes no sense, it doesn't matter. If you read the loss function, you wiil find that. It is a remaining issue of old code. I will modify this problem in the near future.

You will find the test is OK.

[LXYTSOS]

Thank you very much, I'll try it.

[LXYTSOS]

I don't quite understand the code:gallery_feature, gallery_labels = query_feature, query_labels = features, labels and then you calculate the similarity between query_feature and gallery_feature.
and can you explain the Recall_at_ks() function in evaluations/recall_at_k.py, especially what's ks_dict stands for?

[bnu-wangxun]

Please ref to the paper below:
[3][H. Oh Song, Y. Xiang, S. Jegelka, and S. Savarese. Deep metric learning via lifted structured feature embedding. In CVPR, 2016.]