6def46044be4f04b45704562396b39108ec12bf0,lib/model/rpn/proposal_target_layer_4.py,_ProposalTargetLayer,_sample_rois_pytorch,#_ProposalTargetLayer#Any#Any#Any#Any#Any#,117

Before Change

                // sampling bg
                bg_rois_per_this_image = rois_per_image - fg_rois_per_this_image
                rand_num = torch.floor(torch.rand(bg_rois_per_this_image).type_as(all_rois)
                                         * bg_num_rois).long()
                bg_inds = bg_inds[rand_num]
            elif fg_num_rois > 0 and bg_num_rois == 0:
                // sampling fg
                rand_num = torch.floor(torch.rand(rois_per_image).type_as(all_rois)

After Change

                // Seems torch.rand has a bug, it will generate very large number and make an error. 
                // We use numpy rand instead. 
                //rand_num = (torch.rand(bg_rois_per_this_image) * bg_num_rois).long().cuda()
                rand_num = np.floor(np.random.rand(bg_rois_per_this_image) * bg_num_rois)
                rand_num = torch.from_numpy(rand_num).long().cuda()
                bg_inds = bg_inds[rand_num]

            elif fg_num_rois > 0 and bg_num_rois == 0:
                // sampling fg

Instances