1

Has anyone implement the Faster-RCNN for TensorFlow version on Windows? I found some related repos as following:

1.Faster-RCNN for TensorFlow on Linux https://github.com/endernewton/tf-faster-rcnn

2.Faster-RCNN for Caffe on Linux https://github.com/rbgirshick/py-faster-rcnn

3.Faster-RCNN for Caffe on Windows https://github.com/MrGF/py-faster-rcnn-windows

I successfully compiled the 'cpu_nms',but encountered error when trying to run demo.py:

tensorflow.python.framework.error_impl.InvalidArgumentError:ValueError：Buffer dtype mismatch, expected 'int_t' but got 'long long'

PS.Didn't compile gpu_nms cuz I don't know how to deal with 'kernel.cu' and 'gpu_nms.pyx'.I tried to do like what https://github.com/MrGF/py-faster-rcnn-windows did on 'setup_cuda.py' but failed, an error exactly the same as https://github.com/MrGF/py-faster-rcnn-windows/issues/17 happened:

LINK : fatal error LNK1181: cannot open input file 'ID=2.obj'

Has anyone implement the Faster-RCNN for TensorFlow version on Windows or can anyone give me some advice?

Thanks a lot.

Yi Shen
  • 36
  • 3
  • Check [this answer](https://stackoverflow.com/a/44562256/7456923), it links to the official TF object detection API, which includes Faster-RCNN and is compatible with windows, does that answer your question ? – gdelab Jan 05 '18 at 09:11

1 Answers1

0
LINK : fatal error LNK1181: cannot open input file 'ID=2.obj'

This error comes from the link command. In the command string, there's a segment 'ID=2.obj'. You can fix it by removing the string segment.

bozhou
  • 1
  • 1