def test_warp_car(): flow = M.KITTI.open_flow(TEST_PNG16_FILE) car = M.open_png16('./data/kitti_car/c_11.png') flow = flow.reshape([1, *flow.shape]) car = car.reshape([1, *car.shape]) car_bar = M.warp(car, flow[..., 0], flow[..., 1], True) with tf.Session() as sess: car_bar = sess.run(car_bar)[0] car_bar = car_bar.astype('uint8')
def test_warp_chair(): flow = M.open_flo(TEST_FLO_FILE) img1 = Image.open('./data/flying_chair/pair/0000/img1.png') ch0 = np.array(img1).astype('float32') flow = flow.reshape([1, *flow.shape]) ch0 = np.expand_dims(ch0, 0) ch1 = M.warp(ch0, flow[..., 0], flow[..., 1], True) with tf.Session() as sess: ch1 = sess.run(ch1)[0] ch1 = ch1.astype('uint8')
def test_grid(): G = np.meshgrid(range(5), range(5)) G = np.stack(G, -1) G_bar = M._grid(5, 5)[0] with tf.Session() as sess: G_bar = sess.run(G_bar) assert np.all(G == G_bar.transpose([1, 0, 2]))
def test_sample(): G = np.meshgrid(range(5), range(5)) G = np.stack(G, -1) G = np.expand_dims(G, 0) G.transpose([0, 2, 1, 3]) X = np.random.rand(1, 5, 5, 3).astype('float32') X_bar = M._sample(X, G[..., 0], G[..., 1]) with tf.Session() as sess: X_bar = sess.run(X_bar) assert np.all(X == X_bar)
def test_open_png16(): X = M.open_png16(TEST_PNG16_FILE) assert X.shape == (375, 1242, 3)
def test_open_flo(): X = M.open_flo(TEST_FLO_FILE) assert X.shape == (384, 512, 2)