gpu_nms does not work.

[zxGuo]

Recently I reinstalled caffe and py-faster-rcnn, everythong works well except that I get a lot of overlapped detection bounding boxes in my result. Then I find it is because the gpu_nms always outputs all the detection results without doing anything, while cpu_nms works well.

Then I tried to track the problem and I found in 'nms_kernel.cu', the function 'nms_kernel' is skipped.

_ global _ void nms_kernel(const int n_boxes, const float nms_overlap_thresh, const float *dev_boxes, unsigned long long *dev_mask)

Since I do not know much about cuda programming, I can not go further. Can someone help me with this problem? Thanks in advance.

[zxGuo]

My environment is RTX 2080, ubuntu 18.04, python 3.7, cuda 10.0

[RizhaoCai]

The

My environment is RTX 2080, ubuntu 18.04, python 3.7, cuda 10.0

The problems maybe because of python3.7. python3.5 may help.

Btw, do you manage to solve the problem with python3.7