clovaai / eresfd Goto Github PK

bbox_utils.py:331: UserWarning: An output with one or more elements was resized since it had shape [37], which does not match the required output shape [34]. This behavior is deprecated, and in a future PyTorch release outputs will not be resized unless they have zero elements. You can explicitly reuse an out tensor t by resizing it, inplace, to zero elements with t.resize_(0). (Triggered internally at  ..\aten\src\ATen\native\Resize.cpp:24.)

        t1 = time.time()

        # NOTE: modified
        # det0 shape: 750, 5
        det0 = detect_face(net, img, shrink, thresh=thresh)

        # multi-scale
        if not single_scale_test:
            print('det1')
            det1 = flip_test(net, img, shrink, thresh=thresh)  # flip test

            print('det2')
            [det2, det3] = multi_scale_test(net, img, max_im_shrink, thresh=thresh)

            det = np.row_stack((det0, det1, det2, det3))
            dets = bbox_vote(det)
        else:
         
            dets = det0

        t2 = time.time()

        print("Detect %04d th image costs %.4f sec" % (counter, t2 - t1))