the way of ResNet-50-Basel
JoyLuo opened this issue · 2 comments
hi @layumi
your paper and code were read carefully. and the ResNet-50-Basel in your paper is higher then others. the question is how to train the baseline. Is the resnet50 with softmaxloss finetuned in the market1501?
and ours is finetune the resnet50 with siamese loss(two losses) using the market1501?
the last question is that the training solver between ResNet-50-Basel and ours are same or not?
thanks for your kind
luo ze
I only use the simple fine-tuning method with classification loss (predict 751 classes).
There are two little tricks, which may help you:
-
Do not fix the parameters in ResNet. Market1501 or CUHK03 are different from ImageNet.
So a little change yields a better result. -
Add dropout layer before the final fc layer. For example, 2048->dropout->2048->751
I found dropout rate =0.75 is great.