sanghoon / pva-faster-rcnn

Demo code for PVANet
https://arxiv.org/abs/1611.08588
Other
650 stars 241 forks source link

Trying a bigger batch size #80

Open bhushangawde opened 7 years ago

bhushangawde commented 7 years ago

Hello @sanghoon I am not able to fix this error!!!! I am trying a batch size of 4 instead of 1! I know the problem is mismatch of rpn_cls_score , rpn_data and rpn_loss_cls!! What should i do to change the layer configurations so that it will adapt to this!

OUTPUT-->

I0609 07:27:52.402818 18589 net.cpp:157] Top shape: 4 168 54 90 (3265920) I0609 07:27:52.402824 18589 net.cpp:157] Top shape: 4 168 54 90 (3265920) I0609 07:27:52.402829 18589 net.cpp:165] Memory required for data: 9355150164 I0609 07:27:52.402835 18589 layer_factory.hpp:77] Creating layer rpn_cls_score_reshape I0609 07:27:52.402853 18589 net.cpp:100] Creating Layer rpn_cls_score_reshape I0609 07:27:52.402859 18589 net.cpp:434] rpn_cls_score_reshape <- rpn_cls_score_rpn_cls_score_0_split_0 I0609 07:27:52.402868 18589 net.cpp:408] rpn_cls_score_reshape -> rpn_cls_score_reshape I0609 07:27:52.402904 18589 net.cpp:150] Setting up rpn_cls_score_reshape I0609 07:27:52.402911 18589 net.cpp:157] Top shape: 4 2 2268 90 (1632960) I0609 07:27:52.402916 18589 net.cpp:165] Memory required for data: 9361682004 I0609 07:27:52.402921 18589 layer_factory.hpp:77] Creating layer rpn_cls_score_reshape_rpn_cls_score_reshape_0_split I0609 07:27:52.402928 18589 net.cpp:100] Creating Layer rpn_cls_score_reshape_rpn_cls_score_reshape_0_split I0609 07:27:52.402935 18589 net.cpp:434] rpn_cls_score_reshape_rpn_cls_score_reshape_0_split <- rpn_cls_score_reshape I0609 07:27:52.402941 18589 net.cpp:408] rpn_cls_score_reshape_rpn_cls_score_reshape_0_split -> rpn_cls_score_reshape_rpn_cls_score_reshape_0_split_0 I0609 07:27:52.402951 18589 net.cpp:408] rpn_cls_score_reshape_rpn_cls_score_reshape_0_split -> rpn_cls_score_reshape_rpn_cls_score_reshape_0_split_1 I0609 07:27:52.402997 18589 net.cpp:150] Setting up rpn_cls_score_reshape_rpn_cls_score_reshape_0_split I0609 07:27:52.403004 18589 net.cpp:157] Top shape: 4 2 2268 90 (1632960) I0609 07:27:52.403009 18589 net.cpp:157] Top shape: 4 2 2268 90 (1632960) I0609 07:27:52.403014 18589 net.cpp:165] Memory required for data: 9374745684 I0609 07:27:52.403019 18589 layer_factory.hpp:77] Creating layer rpn-data I0609 07:27:52.403735 18589 net.cpp:100] Creating Layer rpn-data I0609 07:27:52.403753 18589 net.cpp:434] rpn-data <- rpn_cls_score_rpn_cls_score_0_split_1 I0609 07:27:52.403760 18589 net.cpp:434] rpn-data <- gt_boxes_input-data_2_split_0 I0609 07:27:52.403767 18589 net.cpp:434] rpn-data <- im_info_input-data_1_split_0 I0609 07:27:52.403774 18589 net.cpp:434] rpn-data <- data_input-data_0_split_1 I0609 07:27:52.403781 18589 net.cpp:408] rpn-data -> rpn_labels I0609 07:27:52.403791 18589 net.cpp:408] rpn-data -> rpn_bbox_targets I0609 07:27:52.403800 18589 net.cpp:408] rpn-data -> rpn_bbox_inside_weights I0609 07:27:52.403808 18589 net.cpp:408] rpn-data -> rpn_bbox_outside_weights I0609 07:27:52.408412 18589 net.cpp:150] Setting up rpn-data I0609 07:27:52.408437 18589 net.cpp:157] Top shape: 1 1 2268 90 (204120) I0609 07:27:52.408444 18589 net.cpp:157] Top shape: 1 168 54 90 (816480) I0609 07:27:52.408450 18589 net.cpp:157] Top shape: 1 168 54 90 (816480) I0609 07:27:52.408457 18589 net.cpp:157] Top shape: 1 168 54 90 (816480) I0609 07:27:52.408463 18589 net.cpp:165] Memory required for data: 9385359924 I0609 07:27:52.408468 18589 layer_factory.hpp:77] Creating layer rpn_loss_cls I0609 07:27:52.408488 18589 net.cpp:100] Creating Layer rpn_loss_cls I0609 07:27:52.408495 18589 net.cpp:434] rpn_loss_cls <- rpn_cls_score_reshape_rpn_cls_score_reshape_0_split_0 I0609 07:27:52.408504 18589 net.cpp:434] rpn_loss_cls <- rpn_labels I0609 07:27:52.408512 18589 net.cpp:408] rpn_loss_cls -> rpn_loss_cls I0609 07:27:52.408529 18589 layer_factory.hpp:77] Creating layer rpn_loss_cls F0609 07:27:52.412192 18589 loss_layer.cpp:19] Check failed: bottom[0]->shape(0) == bottom[1]->shape(0) (4 vs. 1) The data and label should have the same first dimension. Check failure stack trace: Aborted