def conv_block_3d()

in utils.py [0:0]

• 36 lines of code
• 10 McCabe index (conditional complexity)

def conv_block_3d(
        inp,
        cweight,
        bweight,
        reuse,
        scope,
        use_stride=True,
        activation=tf.nn.leaky_relu,
        pn=False,
        bn=False,
        gn=False,
        ln=False,
        scale=None,
        bias=None,
        use_bias=False):
    """ Perform, conv, batch norm, nonlinearity, and max pool """
    stride, no_stride = [1, 1, 2, 2, 1], [1, 1, 1, 1, 1]
    _, d, h, w, _ = inp.get_shape()

    if not use_bias:
        bweight = 0

    if not use_stride:
        conv_output = tf.nn.conv3d(inp, cweight, no_stride, 'SAME') + bweight
    else:
        conv_output = tf.nn.conv3d(inp, cweight, stride, 'SAME') + bweight

    if activation is not None:
        conv_output = activation(conv_output, alpha=0.1)

    if bn:
        conv_output = batch_norm(conv_output, scale, bias)
    if pn:
        conv_output = pixel_norm(conv_output)
    if gn:
        conv_output = group_norm(conv_output, scale, bias)
    if ln:
        conv_output = layer_norm(conv_output, scale, bias)

    if FLAGS.downsample and use_stride:
        conv_output = tf.layers.average_pooling2d(conv_output, (2, 2), 2)

    return conv_output