MobileNet-YOLO
MobileNet-YOLO copied to clipboard
eric612
怎样使用我的数据训练yolov3
已经修改了train和test prototxt中的class_num 和LMDB的路径，也修改了labelmap_voc.prototxt的分类，运行train_yolov3_darknet.脚本会报错如下： F0307 10:05:11.701001 26093 yolov3_layer.cpp:161] Check failed: input_count == tmp_input_count (92055 vs. 7581) *** Check failure stack trace: *** @ 0x7f18961f95cd google::LogMessage::Fail() @ 0x7f18961fb433 google::LogMessage::SendToLog() @ 0x7f18961f915b google::LogMessage::Flush() @ 0x7f18961fbe1e google::LogMessageFatal::~LogMessageFatal() @ 0x7f189699de5a caffe::Yolov3Layer<>::LayerSetUp() @ 0x7f18967bf67c caffe::Net<>::Init() @ 0x7f18967c1d6e caffe::Net<>::Net() @ 0x7f18969dcaca caffe::Solver<>::InitTrainNet() @ 0x7f18969de095 caffe::Solver<>::Init() @ 0x7f18969de3af caffe::Solver<>::Solver() @ 0x7f1896800af1 caffe::Creator_SGDSolver<>() @ 0x40df4a train() @ 0x40afd7 main @ 0x7f189569b830 __libc_start_main @ 0x40b979 _start @ (nil) (unknown) Aborted
尝试训练 yolov3-lite，按照wiki中给的链接修改了classes和num_output，但是依旧报错如下： I0307 15:03:12.709883 29872 layer_factory.hpp:77] Creating layer conv12/dw/relu @ 0x7f18c3932865 caffe::InternalThread::entry() I0307 15:03:12.709888 29872 net.cpp:86] Creating Layer conv12/dw/relu I0307 15:03:12.709890 29872 net.cpp:408] conv12/dw/relu <- conv12/dw I0307 15:03:12.709894 29872 net.cpp:369] conv12/dw/relu -> conv12/dw (in-place) I0307 15:03:12.709898 29872 net.cpp:124] Setting up conv12/dw/relu I0307 15:03:12.709903 29872 net.cpp:131] Top shape: 7 512 19 19 (1293824) I0307 15:03:12.709906 29872 net.cpp:139] Memory required for data: 4088609840 I0307 15:03:12.709908 29872 layer_factory.hpp:77] Creating layer conv12 I0307 15:03:12.709913 29872 net.cpp:86] Creating Layer conv12 I0307 15:03:12.709916 29872 net.cpp:408] conv12 <- conv12/dw I0307 15:03:12.709920 29872 net.cpp:382] conv12 -> conv12 @ 0x7f18c393378e boost::detail::thread_data<>::run() @ 0x7f18c25d25d5 (unknown) @ 0x7f18c21936ba start_thread @ 0x7f18c28ee41d clone @ 0x0 (unknown) Aborted
可以參考以下連結 https://github.com/eric612/MobileNet-YOLO/issues/47 https://github.com/PiyalGeorge/MobileNet-YOLO/tree/master/models/yolov3
可以參考以下連結 #47 https://github.com/PiyalGeorge/MobileNet-YOLO/tree/master/models/yolov3
感谢您的回复，我按照您第二个链接中的步骤修改了test和train prototxt，但是运行train_yolov3_lite.sh开始尝试训练的时候就会报错终止，log如下： I0307 15:43:18.103624 31208 net.cpp:408] conv12 <- conv12/dw I0307 15:43:18.103631 31208 net.cpp:382] conv12 -> conv12 @ 0x7f5b22fb3406 caffe::AnnotatedDataLayer<>::load_batch() @ 0x7f5b22fc8c83 caffe::BasePrefetchingDataLayer<>::InternalThreadEntry() @ 0x7f5b22dfd865 caffe::InternalThread::entry() @ 0x7f5b22dfe78e boost::detail::thread_data<>::run() @ 0x7f5b21a9d5d5 (unknown) I0307 15:43:18.107707 31208 net.cpp:124] Setting up conv12 I0307 15:43:18.107720 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.107724 31208 net.cpp:139] Memory required for data: 4098960432 I0307 15:43:18.107728 31208 layer_factory.hpp:77] Creating layer conv12/bn I0307 15:43:18.107734 31208 net.cpp:86] Creating Layer conv12/bn I0307 15:43:18.107738 31208 net.cpp:408] conv12/bn <- conv12 I0307 15:43:18.107743 31208 net.cpp:369] conv12/bn -> conv12 (in-place) I0307 15:43:18.107862 31208 net.cpp:124] Setting up conv12/bn I0307 15:43:18.107867 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.107869 31208 net.cpp:139] Memory required for data: 4109311024 I0307 15:43:18.107874 31208 layer_factory.hpp:77] Creating layer conv12/scale I0307 15:43:18.107879 31208 net.cpp:86] Creating Layer conv12/scale I0307 15:43:18.107882 31208 net.cpp:408] conv12/scale <- conv12 I0307 15:43:18.107897 31208 net.cpp:369] conv12/scale -> conv12 (in-place) I0307 15:43:18.107921 31208 layer_factory.hpp:77] Creating layer conv12/scale I0307 15:43:18.107990 31208 net.cpp:124] Setting up conv12/scale I0307 15:43:18.107995 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.107996 31208 net.cpp:139] Memory required for data: 4119661616 I0307 15:43:18.108001 31208 layer_factory.hpp:77] Creating layer conv12/relu I0307 15:43:18.108005 31208 net.cpp:86] Creating Layer conv12/relu I0307 15:43:18.108006 31208 net.cpp:408] conv12/relu <- conv12 I0307 15:43:18.108011 31208 net.cpp:369] conv12/relu -> conv12 (in-place) I0307 15:43:18.108016 31208 net.cpp:124] Setting up conv12/relu I0307 15:43:18.108018 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.108021 31208 net.cpp:139] Memory required for data: 4130012208 I0307 15:43:18.108023 31208 layer_factory.hpp:77] Creating layer conv13/dw I0307 15:43:18.108028 31208 net.cpp:86] Creating Layer conv13/dw I0307 15:43:18.108031 31208 net.cpp:408] conv13/dw <- conv12 I0307 15:43:18.108036 31208 net.cpp:382] conv13/dw -> conv13/dw I0307 15:43:18.108217 31208 net.cpp:124] Setting up conv13/dw I0307 15:43:18.108223 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.108225 31208 net.cpp:139] Memory required for data: 4140362800 I0307 15:43:18.108229 31208 layer_factory.hpp:77] Creating layer conv13/dw/bn I0307 15:43:18.108233 31208 net.cpp:86] Creating Layer conv13/dw/bn I0307 15:43:18.108237 31208 net.cpp:408] conv13/dw/bn <- conv13/dw I0307 15:43:18.108240 31208 net.cpp:369] conv13/dw/bn -> conv13/dw (in-place) @ 0x7f5b2165e6ba start_thread I0307 15:43:18.108357 31208 net.cpp:124] Setting up conv13/dw/bn I0307 15:43:18.108361 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.108366 31208 net.cpp:139] Memory required for data: 4150713392 I0307 15:43:18.108369 31208 layer_factory.hpp:77] Creating layer conv13/dw/scale I0307 15:43:18.108373 31208 net.cpp:86] Creating Layer conv13/dw/scale I0307 15:43:18.108376 31208 net.cpp:408] conv13/dw/scale <- conv13/dw I0307 15:43:18.108381 31208 net.cpp:369] conv13/dw/scale -> conv13/dw (in-place) I0307 15:43:18.108402 31208 layer_factory.hpp:77] Creating layer conv13/dw/scale I0307 15:43:18.108467 31208 net.cpp:124] Setting up conv13/dw/scale I0307 15:43:18.108472 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.108475 31208 net.cpp:139] Memory required for data: 4161063984 I0307 15:43:18.108479 31208 layer_factory.hpp:77] Creating layer conv13/dw/relu I0307 15:43:18.108482 31208 net.cpp:86] Creating Layer conv13/dw/relu I0307 15:43:18.108485 31208 net.cpp:408] conv13/dw/relu <- conv13/dw I0307 15:43:18.108489 31208 net.cpp:369] conv13/dw/relu -> conv13/dw (in-place) I0307 15:43:18.108492 31208 net.cpp:124] Setting up conv13/dw/relu I0307 15:43:18.108495 31208 net.cpp:131] Top shape: 7 1024 19 19 (2587648) I0307 15:43:18.108508 31208 net.cpp:139] Memory required for data: 4171414576 I0307 15:43:18.108511 31208 layer_factory.hpp:77] Creating layer conv13 I0307 15:43:18.108517 31208 net.cpp:86] Creating Layer conv13 I0307 15:43:18.108520 31208 net.cpp:408] conv13 <- conv13/dw I0307 15:43:18.108525 31208 net.cpp:382] conv13 -> conv13 @ 0x7f5b21db941d clone @ 0x0 (unknown) Aborted
补充一下，我的数据集是曾经用来训练mobilenet-ssd和rfcn的，所以我想应该不是数据集的问题，但是按照您给的链接，我已经都修改好了，不清楚为什么还是无法开始训练。再次感谢您的帮助！
hi,Have you solved the problem?I met the same problem as you.can you help me ?
