faster rcnn support error

guoqiang0148666 commented 5 years ago

Hello, l want to use nvidia caffe to support faster rcnn, but l meet a error as following:

F0121 16:54:30.924830 13972 blob.hpp:593] Check failed: current_type == new_type (3 vs. 1) Attempt to change TBlob native data type from INT to FLOAT Check failure stack trace: @ 0x7f3420fa75cd google::LogMessage::Fail() @ 0x7f3420fa9433 google::LogMessage::SendToLog() @ 0x7f3420fa715b google::LogMessage::Flush() @ 0x7f3420fa9e1e google::LogMessageFatal::~LogMessageFatal() @ 0x7f3421f18b9d caffe::ROIPoolingLayer<>::Forward_gpu() @ 0x7f3422f0e382 caffe::Layer<>::Forward() @ 0x7f3421cdd7f3 caffe::Net::ForwardFromTo() @ 0x7f3421cdd957 caffe::Net::Forward() @ 0x7f3421ce2345 caffe::Net::ForwardBackward() @ 0x7f3421cf8eab caffe::Solver::Step() @ 0x7f3422ef141d caffe::Solver_StepNoGIL() @ 0x7f3422f0d304 boost::python::objects::caller_py_function_impl<>::operator()() @ 0x7f34206d95cd boost::python::objects::function::call() @ 0x7f34206d97c8 (unknown) @ 0x7f34206e1613 boost::python::handle_exception_impl() @ 0x7f34206d6999 (unknown) @ 0x4c166d PyEval_EvalFrameEx @ 0x4c141f PyEval_EvalFrameEx @ 0x4b9b66 PyEval_EvalCodeEx @ 0x4c17c6 PyEval_EvalFrameEx @ 0x4b9b66 PyEval_EvalCodeEx @ 0x4eb69f (unknown) @ 0x4e58f2 PyRun_FileExFlags @ 0x4e41a6 PyRun_SimpleFileExFlags @ 0x4938ce Py_Main @ 0x7f345859a830 __libc_start_main @ 0x493299 _start @ (nil) (unknown) Aborted (core dumped)

Any suggestions are welcome, thank you

drnikolaev commented 5 years ago

@guoqiang0148666 what is ROIPoolingLayer?

guoqiang0148666 commented 5 years ago

ROIPoolingLayer comes from faster rcnn (https://github.com/rbgirshick/py-faster-rcnn)

whria78 commented 5 years ago

1) ROIPooling Layer was supported by Microsoft version of Caffe (2015).

:~/caffe/second/py-faster-rcnn/caffe-fast-rcnn$ git log
commit 0dcd397b29507b8314e252e850518c5695efbb83
Author: Ross Girshick <ross.girshick@gmail.com>
Date:   Sun Aug 9 13:41:39 2015 -0700

    Fast and Faster R-CNN change set
     - smooth l1 loss
     - roi pooling
     - expose phase in pycaffe
     - dropout scaling at test time (needed for MSRA-trained ZF network)

src/caffe/proto/caffe.proto
optional ROIPoolingParameter roi_pooling_param = 8266711;

2) py-faster-rcnn works well on Ubuntu 16.04 (https://github.com/rbgirshick/py-faster-rcnn). However, if we want to use it with the lastest cuDNN on Ubuntu 18.04, please check this post (https://gist.github.com/acmiyaguchi/bc535ba23eabd3564edd73e491763d50)

drnikolaev commented 5 years ago

This is about BVLC Caffe, closing

NVIDIA / caffe

faster rcnn support error #553