コード例 #1
0
    def test_normalize_homography_identity(self, batch_size, device, dtype):
        # create input data
        input_shape = (4, 8, 5)
        dst_homo_src = utils.create_eye_batch(batch_size=batch_size, eye_size=4).to(device=device, dtype=dtype)

        res = torch.tensor(
            [[[0.5000, 0.0, 0.0, -1.0], [0.0, 0.2857, 0.0, -1.0], [0.0, 0.0, 0.6667, -1.0], [0.0, 0.0, 0.0, 1.0]]],
            device=device,
            dtype=dtype,
        )
        norm = kornia.normal_transform_pixel3d(input_shape[0], input_shape[1], input_shape[2]).to(
            device=device, dtype=dtype
        )
        assert_close(norm, res, rtol=1e-4, atol=1e-4)

        norm_homo = kornia.normalize_homography3d(dst_homo_src, input_shape, input_shape).to(device=device, dtype=dtype)
        assert_close(norm_homo, dst_homo_src, rtol=1e-4, atol=1e-4)

        norm_homo = kornia.normalize_homography3d(dst_homo_src, input_shape, input_shape).to(device=device, dtype=dtype)
        assert_close(norm_homo, dst_homo_src, rtol=1e-4, atol=1e-4)

        # change output scale
        norm_homo = kornia.normalize_homography3d(
            dst_homo_src, input_shape, (input_shape[0] // 2, input_shape[1] * 2, input_shape[2] // 2)
        ).to(device=device, dtype=dtype)
        res = torch.tensor(
            [[[4.0, 0.0, 0.0, 3.0], [0.0, 0.4667, 0.0, -0.5333], [0.0, 0.0, 3.0, 2.0], [0.0, 0.0, 0.0, 1.0]]],
            device=device,
            dtype=dtype,
        ).repeat(batch_size, 1, 1)
        assert_close(norm_homo, res, rtol=1e-4, atol=1e-4)
コード例 #2
0
    def test_normalize_homography_general(self, batch_size, device, dtype):
        # create input data
        dst_homo_src = torch.eye(4, device=device, dtype=dtype)
        dst_homo_src[..., 0, 0] = 0.5
        dst_homo_src[..., 1, 1] = 0.5
        dst_homo_src[..., 2, 2] = 2.0
        dst_homo_src[..., 0, 3] = 1.0
        dst_homo_src[..., 1, 3] = 2.0
        dst_homo_src[..., 2, 3] = 3.0
        dst_homo_src = dst_homo_src.expand(batch_size, -1, -1)

        norm_homo = kornia.normalize_homography3d(dst_homo_src, (2, 2, 5), (2, 2, 5))
        res = torch.tensor([[[0.5, 0.0, 0.0, 0.0],
                             [0.0, 0.5, 0.0, 3.5],
                             [0.0, 0.0, 2.0, 7.0],
                             [0.0, 0.0, 0.0, 1.0]]], device=device, dtype=dtype)
        assert (norm_homo == res).all()