Fine_Grained_Classification icon indicating copy to clipboard operation
Fine_Grained_Classification copied to clipboard

The Rank_Loss in train_val_fixcls.prototxt?

Open RTMDFG opened this issue 6 years ago • 5 comments

Hi ,guy I have a question about the rank_loss in train_val_fixcls.prototxt.In the RA-CNN/alternate_training/train_val_fixcls.prototxt, you concat prob1,prob2,prob3 compute rank_loss.I think there are two rank_loss.The first rank_loss between prob1,prob2 and the second rank_loss between prob2,prob3. @ouceduxzk

RTMDFG avatar Mar 15 '18 08:03 RTMDFG

@RTMDFG @ouceduxzk how are you going to run the model under alternate optimisation settings (APN vs CLS)?

abhimahesh avatar Mar 17 '18 18:03 abhimahesh

@abhimahesh I train the Network.But the train_val.prototxt is unlike the ouceduxzk . I set two rank_loss.

RTMDFG avatar Mar 18 '18 05:03 RTMDFG

@RTMDFG are you getting similar results as mentioned in paper? Can you share your code ? Whats your e-mail ID?

abhimahesh avatar Mar 18 '18 07:03 abhimahesh

@RTMDFG you are probably right, I did not have time to develop further, currently the APN pretrained net is not finished yet, so please contribute @abhimahesh , not just asking for sharing the code

ouceduxzk avatar Mar 18 '18 11:03 ouceduxzk

@abhimahesh I don't train it in the bird dataset.So i don't acheive the results. You can reference here (https://gist.github.com/jens25/6b0ea1143599fb99bd499a08dd5c072c).

RTMDFG avatar Mar 21 '18 01:03 RTMDFG