Is anybody interested in training new model using faster rcnn?

[X.T. Li]

I found a python version of faster rcnn, but I want to use it on the imagenet detection data set, is anyone interested in this?

[Alex Bewley]

I would be interested in looking at it. Can you please send a link to the python version?

[X.T. Li]

Here is the link: http://github.com.proxy.parle.co/rbgirshick/py-faster-rcnn

在 2015年10月15日星期四 UTC+8上午5:06:48，Alex Bewley写道：

[waker]

I am focusing on faster rcnn and I want to use it on specific object detection.  I will follow any progress on it

在 2015年10月14日星期三 UTC+8上午9:50:14，X.T. Li写道：

[Alex Bewley]

I finally got faster-rcnn to work on my mac (only in cpu mode so far). However, the demo appears to be nondeterministic. I can't quite work out what is the cause. Is anyone else experiences similar issues? I.e. Run it once and it detects most objects and run it a second time without any changes and it will either miss objects or appears to not perform NMS. Similar behavior occurs with both ZF and VGG16 nets.

[Indhu TR]

Can you pls tell how to train FRCNN in CPU mode? I found on one link to change the file roi_pooling_layers.cpp and smmoth_L1_loss_layers.cpp. I replced those files with new ones and after that i gave make && make pycaffe. But i am getting the following error

LD -o .build_release/lib/libcaffe.so.1.0.0-rc3

/usr/bin/ld: cannot find -lopencv_imgcodecs

/usr/bin/ld: cannot find -lopenblas

collect2: error: ld returned 1 exit status

make: *** [.build_release/lib/libcaffe.so.1.0.0-rc3] Error 1

Any help?

[abhigoku10]

perform the following ##Line 89
	# in the makefile add opencv_imgcodecs into the LIBRARIES line and include the opencv
	LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu /usr/lib/x86_64-linux-gnu/hdf5/serial /home/adminvatm/opencv/build/lib
	INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial/