cannot run train_detection.sh

[wenhe-jia]

When I try to run you training script to train a vgg cascade faster rcnn, I encounter a error like this:

I0426 15:14:21.523888 14111 layer_factory.hpp:77] Creating layer data F0426 15:14:21.523939 14111 layer_factory.hpp:81] Check failed: registry.count(type) == 1 (0 vs. 1) Unknown layer type: DetectionData (known types: AbsVal, Accuracy, ArgMax, BNLL, BatchNorm, BatchReindex, BboxAccuracy, Bias, BoxGroupOutput, Concat, ContrastiveLoss, Convolution, Crop, Data, DecodeBBox, Deconvolution, DetectionEvaluate, DetectionGroupAccuracy, DetectionGroupLoss, Dropout, DummyData, ELU, Eltwise, Embed, EuclideanLoss, Exp, Filter, Flatten, FrcnnOutput, HDF5Data, HDF5Output, HardMining, HingeLoss, Im2col, InfogainLoss, InnerProduct, Input, LRN, LSTM, LSTMUnit, Log, MVN, MemoryData, MultinomialLogisticLoss, PReLU, PSROIPooling, Parameter, Pooling, Power, ProposalTarget, Python, RNN, ROIAlign, ROIPooling, ReLU, Reduction, Reshape, SPP, Scale, ScaleRoute, Sigmoid, SigmoidCrossEntropyLoss, Silence, Slice, SmoothL1Loss, Softmax, SoftmaxWithLoss, Split, TanH, Threshold, Tile) Check failure stack trace: @ 0x7f87d5321daa (unknown) @ 0x7f87d5321ce4 (unknown) @ 0x7f87d53216e6 (unknown) @ 0x7f87d5324687 (unknown) @ 0x7f87d59a7f89 caffe::LayerRegistry<>::CreateLayer() @ 0x7f87d59aa488 caffe::Net<>::Init() @ 0x7f87d59ac262 caffe::Net<>::Net() @ 0x7f87d5b3eab0 caffe::Solver<>::InitTrainNet() @ 0x7f87d5b3efc3 caffe::Solver<>::Init() @ 0x7f87d5b3f27f caffe::Solver<>::Solver() @ 0x7f87d5b31241 caffe::Creator_SGDSolver<>() @ 0x414437 caffe::SolverRegistry<>::CreateSolver() @ 0x40dbd5 train() @ 0x40a02c main @ 0x7f87d3b81f45 (unknown) @ 0x40aa6b (unknown) @ (nil) (unknown) Aborted (core dumped)

I have checked the caffe.proto and the parameters of _detection_datalayer are already registered, can you give me some advise?

[zhaoweicai]

The problem may be relevant with OpenCV. This whole data layer depends on OpenCV. If you set USE_OPENCV := 0 in the Makefile.config and it will not be complied. Try to turn on the OpenCV.

[wenhe-jia]

I have set USE_OPENCV = 1, in the Makefile.config, and the compilation goes well without warnings and errors. And I am using opencv 3, so I uncomment it in Makefile.config as well.

[zhaoweicai]

Hi @LeonJWH, I tried you Makefile.config, but I can't reproduce your error. Could you check your compilation log if the DetectionData layer was complied?

[wenhe-jia]

thank you！ In the compilation log, the DetectionData layer is compiled, but the problem still exits. I also tried to run your repo on another machine, it went well. So I guess is my machine's problem, sorry to bother you.