Faster RCNN训练出现问题:smooth_L1_loss_layer.cpp:28] Check failed: bottom[0]->channels() == bottom[1]->cha
2017-10-30 21:31
609 查看
I1030 20:49:01.633868 24921 net.cpp:157] Top shape: 1 4096 (4096) I1030 20:49:01.633885 24921 net.cpp:165] Memory required for data: 280497656 I1030 20:49:01.633891 24921 layer_factory.hpp:77] Creating layer relu7 I1030 20:49:01.633898 24921 net.cpp:106] Creating Layer relu7 I1030 20:49:01.633900 24921 net.cpp:454] relu7 <- fc7 I1030 20:49:01.633905 24921 net.cpp:397] relu7 -> fc7 (in-place) I1030 20:49:01.634315 24921 net.cpp:150] Setting up relu7 I1030 20:49:01.634320 24921 net.cpp:157] Top shape: 1 4096 (4096) I1030 20:49:01.634336 24921 net.cpp:165] Memory required for data: 280514040 I1030 20:49:01.634338 24921 layer_factory.hpp:77] Creating layer drop7 I1030 20:49:01.634343 24921 net.cpp:106] Creating Layer drop7 I1030 20:49:01.634346 24921 net.cpp:454] drop7 <- fc7 I1030 20:49:01.634348 24921 net.cpp:397] drop7 -> fc7 (in-place) I1030 20:49:01.634369 24921 net.cpp:150] Setting up drop7 I1030 20:49:01.634372 24921 net.cpp:157] Top shape: 1 4096 (4096) I1030 20:49:01.634373 24921 net.cpp:165] Memory required for data: 280530424 I1030 20:49:01.634376 24921 layer_factory.hpp:77] Creating layer fc7_drop7_0_split I1030 20:49:01.634378 24921 net.cpp:106] Creating Layer fc7_drop7_0_split I1030 20:49:01.634379 24921 net.cpp:454] fc7_drop7_0_split <- fc7 I1030 20:49:01.634383 24921 net.cpp:411] fc7_drop7_0_split -> fc7_drop7_0_split_0 I1030 20:49:01.634387 24921 net.cpp:411] fc7_drop7_0_split -> fc7_drop7_0_split_1 I1030 20:49:01.634412 24921 net.cpp:150] Setting up fc7_drop7_0_split I1030 20:49:01.634414 24921 net.cpp:157] Top shape: 1 4096 (4096) I1030 20:49:01.634416 24921 net.cpp:157] Top shape: 1 4096 (4096) I1030 20:49:01.634418 24921 net.cpp:165] Memory required for data: 280563192 I1030 20:49:01.634418 24921 layer_factory.hpp:77] Creating layer cls_score I1030 20:49:01.634423 24921 net.cpp:106] Creating Layer cls_score I1030 20:49:01.634424 24921 net.cpp:454] cls_score <- fc7_drop7_0_split_0 I1030 20:49:01.634428 24921 net.cpp:411] cls_score -> cls_score I1030 20:49:01.634549 24921 net.cpp:150] Setting up cls_score I1030 20:49:01.634552 24921 net.cpp:157] Top shape: 1 2 (2) I1030 20:49:01.634553 24921 net.cpp:165] Memory required for data: 280563200 I1030 20:49:01.634557 24921 layer_factory.hpp:77] Creating layer bbox_pred I1030 20:49:01.634574 24921 net.cpp:106] Creating Layer bbox_pred I1030 20:49:01.634575 24921 net.cpp:454] bbox_pred <- fc7_drop7_0_split_1 I1030 20:49:01.634578 24921 net.cpp:411] bbox_pred -> bbox_pred I1030 20:49:01.634840 24921 net.cpp:150] Setting up bbox_pred I1030 20:49:01.634843 24921 net.cpp:157] Top shape: 1 8 (8) I1030 20:49:01.634845 24921 net.cpp:165] Memory required for data: 280563232 I1030 20:49:01.634861 24921 layer_factory.hpp:77] Creating layer loss_cls I1030 20:49:01.634865 24921 net.cpp:106] Creating Layer loss_cls I1030 20:49:01.634867 24921 net.cpp:454] loss_cls <- cls_score I1030 20:49:01.634869 24921 net.cpp:454] loss_cls <- labels I1030 20:49:01.634872 24921 net.cpp:411] loss_cls -> cls_loss I1030 20:49:01.634876 24921 layer_factory.hpp:77] Creating layer loss_cls I1030 20:49:01.635040 24921 net.cpp:150] Setting up loss_cls I1030 20:49:01.635044 24921 net.cpp:157] Top shape: (1) I1030 20:49:01.635046 24921 net.cpp:160] with loss weight 1 I1030 20:49:01.635066 24921 net.cpp:165] Memory required for data: 280563236 I1030 20:49:01.635067 24921 layer_factory.hpp:77] Creating layer loss_bbox I1030 20:49:01.635071 24921 net.cpp:106] Creating Layer loss_bbox I1030 20:49:01.635073 24921 net.cpp:454] loss_bbox <- bbox_pred I1030 20:49:01.635076 24921 net.cpp:454] loss_bbox <- bbox_targets I1030 20:49:01.635078 24921 net.cpp:454] loss_bbox <- bbox_inside_weights I1030 20:49:01.635080 24921 net.cpp:454] loss_bbox <- bbox_outside_weights I1030 20:49:01.635083 24921 net.cpp:411] loss_bbox -> bbox_loss F1030 20:49:01.635092 24921 smooth_L1_loss_layer.cpp:28] Check failed: bottom[0]->channels() == bottom[1]->channels() (8 vs. 84) *** Check failure stack trace: *** ./experiments/scripts/faster_rcnn_end2end.sh: 行 57: 24921 已放弃 (核心已转储) ./tools/train_net.py --gpu ${GPU_ID} --solver models/${PT_DIR}/${NET}/faster_rcnn_end2end/solver.prototxt --weights data/imagenet_models/${NET}.v2.caffemodel --imdb ${TRAIN_IMDB} --iters ${ITERS} --cfg experiments/cfgs/faster_rcnn_end2end.yml ${EXTRA_ARGS}
修改:
layer { name: 'roi-data' type: 'Python' bottom: 'rpn_rois' bottom: 'gt_boxes' top: 'rois' top: 'labels' top: 'bbox_targets' top: 'bbox_inside_weights' top: 'bbox_outside_weights' python_param { module: 'rpn.proposal_target_layer' layer: 'ProposalTargetLayer' param_str: "'num_classes': 2" }
相关文章推荐
- Faster RCNN训练出现问题:generated_database_->Add(encoded_file_descriptor, size)
- Faster RCNN训练出现问题:ValueError: operands could not be broadcast together with shapes
- Faster RCNN训练出现问题:Matlab版本训练
- faster rcnn 如何训练voc2007数据集 及出现的小问题
- faster rcnn训练过程出现loss=nan解决办法
- Faster RCNN训练出现问题:Selective search data not found at: /home/py-faster-rcnn/data/selective_search_dat
- 解决faster-rcnn中训练时assert(boxes[:,2]>=boxes[:,0]).all()的问题
- 解决faster-rcnn中训练时assert(boxes[:,2]>=boxes[:,0]).all()的问题
- faster-rcnn系列assert (boxes[:, 2] >= boxes[:, 0]).all()和loss偶尔为nan的问题
- smooth_L1_loss_layer.cu解读 caffe源码初认识
- 解决faster-rcnn中训练时assert(boxes[:,2]>=boxes[:,0]).all()的问题
- faster r-cnn训练问题集
- 解决faster-rcnn中训练时assert(boxes[:,2]>=boxes[:,0]).all()的问题
- 训练Faster R-CNN出现的errors
- 解决faster-rcnn中训练时assert(boxes[:,2]>=boxes[:,0]).all()的问题
- Faster R-CNN训练问题解决
- 解决faster-rcnn中训练时assert(boxes[:,2]>=boxes[:,0]).all()的问题
- eclipse调用caffe后py-faster-rcnn出现的问题几解决方法(cuda8.0+cudnn.V5 )
- 解决faster-rcnn中训练时assert(boxes[:,2]>=boxes[:,0]).all()的问题
- faster-rcnn系列assert (boxes[:, 2] >= boxes[:, 0]).all()和loss偶尔为nan的问题