sanghoon
commented
7 years ago Thank you for your report. I'm currently on NIPS2016 to present our work at a workshop tomorrow. I'll definitely check it when I get back to my home.
Open xiaoxiongli opened 7 years ago
sanghoon
commented
7 years ago Thank you for your report. I'm currently on NIPS2016 to present our work at a workshop tomorrow. I'll definitely check it when I get back to my home.
xiaoxiongli
commented
7 years ago WOW! NIPS2016!!! nice~ i have never been to such a big conference... , wish you have a wonderful workshop and travelling
Dear sanghoon: ^_^ @sanghoon I found a bug(i am not sure) list below:
in file test.py: im_scale_factors.append(np.array([im_scale_x, im_scale_y, im_scale_x, im_scale_y]))
blobs['im_info'] = np.array( [np.hstack((im_blob.shape[2], im_blob.shape[3], im_scales[0]))], dtype=np.float32)
which means that blobs['im_info'] = [height(0), width(1), im_scale_x(2), im_scale_y(3), im_scale_x(4), im_scale_y(5)], "(number)" stand for the array's index started from zero
in file proposal_layer.cpp's Forward_cpu(maybe Forward_gpu also have this issue): // input image height & width const Dtype img_H = p_img_info_cpu[0]; const Dtype img_W = p_img_info_cpu[1]; // scale factor for height & width const Dtype scale_H = p_img_info_cpu[2]; <---- it seems here should be p_img_info_cpu[3] const Dtype scale_W = p_img_info_cpu[3]; <---- it seems here should be p_img_info_cpu[2]
Although this is a bug, but it seems affect not too much, because scale_H and scale_W's value is always similar, and it seems scale_H and scale_W are only affect the minimal size（min_box_H、min_box_W） of the bounding box. i wonder is it a bug? ......