Usage on CPU

I would also like to use Faster RCNN without GPUs. Currently when trying to compile this on a non GPU machine I get the following error:

Traceback (most recent call last):
  File "setup.py", line 55, in <module>
    CUDA = locate_cuda()
  File "setup.py", line 43, in locate_cuda
    raise EnvironmentError('The nvcc binary could not be '
EnvironmentError: The nvcc binary could not be located in your $PATH. Either add it to your path, or set $CUDAHOME
make: *** [all] Error 1

Does anyone have experience on how to make this implementation work without GPUs? Code wise it should probably be possible, I guess it is just adapting the build somehow?

[shuang1330]

It might be too late for you, but maybe useful for other people… 1) Change the setup.py file by replacing all ‘raise EnvironmentError’ with ‘return None’, disabling the Extension(‘nms.gpu_nms’…). 2) Change the lib/model/nms_wrapper.py file by disabling ‘import gpu_nms’ and setting ‘force_cpu’ to be True. Above works for me at least.

[FrankXinqi]

I am getting an error “‘MSVCCompiler’ object has no attribute ‘compiler_so’”, when I am running on Windows. How can I solve it?

Same Problem…