def test_flip_regression():
coords = np.random.rand(3, 3)
flip_pairs = [[1, 2]]
root = coords[:1]
coords_flipped = coords.copy()
coords_flipped[1] = coords[2]
coords_flipped[2] = coords[1]
coords_flipped[..., 0] = 2 * root[..., 0] - coords_flipped[..., 0]
# static mode
res_static = fliplr_regression(coords,
flip_pairs,
center_mode='static',
center_x=root[0, 0])
assert_array_almost_equal(res_static, coords_flipped)
# root mode
res_root = fliplr_regression(coords,
flip_pairs,
center_mode='root',
center_index=0)
assert_array_almost_equal(res_root, coords_flipped)
def inference_model(self, x, flip_pairs=None):
"""Inference function.
Returns:
output_regression (np.ndarray): Output regression.
Args:
x (torch.Tensor[N, K, 2]): Input features.
flip_pairs (None | list[tuple()):
Pairs of keypoints which are mirrored.
"""
output = self.forward(x)
if flip_pairs is not None:
output_regression = fliplr_regression(
output.detach().cpu().numpy(),
flip_pairs,
center_mode='static',
center_x=0)
else:
output_regression = output.detach().cpu().numpy()
return output_regression
