SoftmaxWithLoss Layer takes 2 bottom blob(s) as input.

[Kaze89]

Hi,
I am trying to run the train.sh in 4_CasFRCN with data output from former 3 steps, but got the error like this:

I0524 22:45:28.129767  8106 layer_factory.hpp:77] Creating layer data
I0524 22:45:28.132279  8106 net.cpp:106] Creating Layer data
I0524 22:45:28.132289  8106 net.cpp:411] data -> data
I0524 22:45:28.132310  8106 net.cpp:411] data -> rois
I0524 22:45:28.132314  8106 net.cpp:411] data -> labels
I0524 22:45:28.132318  8106 net.cpp:411] data -> cls_loss_weights
I0524 22:45:28.132323  8106 net.cpp:411] data -> bbox_targets
I0524 22:45:28.132329  8106 net.cpp:411] data -> bbox_loss_weights
I0524 22:45:28.358902  8106 net.cpp:150] Setting up data
...
I0524 22:45:28.550379  8106 layer_factory.hpp:77] Creating layer loss_cls
I0524 22:45:28.550384  8106 net.cpp:106] Creating Layer loss_cls
I0524 22:45:28.550385  8106 net.cpp:454] loss_cls <- cls_score_reshape
I0524 22:45:28.550387  8106 net.cpp:454] loss_cls <- labels_reshape
I0524 22:45:28.550402  8106 net.cpp:454] loss_cls <- cls_loss_weights_reshape
I0524 22:45:28.550407  8106 net.cpp:411] loss_cls -> loss_cls
F0524 22:45:28.550436  8106 layer.hpp:375] Check failed: ExactNumBottomBlobs() == bottom.size() (2 vs. 3) SoftmaxWithLoss Layer takes 2 bottom blob(s) as input.

It seems that SoftmaxWithLoss Layer cannot take bottom blobs other than 2, but this prototxt gives 3.
Now I couldn't solve this problem. Can anybody give suggestions?

[Bin Yang]

Use my modified caffe which adds a weights input to SoftmaxWithLoss layer.