Function pad

examples/OpticalFlow/flownet_models.py:17–31

Pad tensor in H, W Remarks: TensorFlow uses "ceil(input_spatial_shape[i] / strides[i])" rather than explicit padding like Caffe, pyTorch does. Hence, we need to pad here beforehand. Args: x (tf.tensor): incoming tensor p (int, optional): padding for H, W

(x, p=3)

def pad(x, p=3):
    """Pad tensor in H, W

    Remarks:
        TensorFlow uses "ceil(input_spatial_shape[i] / strides[i])" rather than explicit padding
        like Caffe, pyTorch does. Hence, we need to pad here beforehand.

    Args:
        x (tf.tensor): incoming tensor
        p (int, optional): padding for H, W

    Returns:
        tf.tensor: padded tensor
    """
    return tf.pad(x, [[0, 0], [0, 0], [p, p], [p, p]])


def channel_norm(x):