(x, scale=1, channel=32, is_training=True,
name='discriminator', patch=True, reuse=False)
| 99 | |
| 100 | |
| 101 | def disc_bn(x, scale=1, channel=32, is_training=True, |
| 102 | name='discriminator', patch=True, reuse=False): |
| 103 | |
| 104 | with tf.variable_scope(name, reuse=reuse): |
| 105 | |
| 106 | for idx in range(3): |
| 107 | x = slim.convolution2d(x, channel*2**idx, [3, 3], stride=2, activation_fn=None) |
| 108 | x = slim.batch_norm(x, is_training=is_training, center=True, scale=True) |
| 109 | x = tf.nn.leaky_relu(x) |
| 110 | |
| 111 | x = slim.convolution2d(x, channel*2**idx, [3, 3], activation_fn=None) |
| 112 | x = slim.batch_norm(x, is_training=is_training, center=True, scale=True) |
| 113 | x = tf.nn.leaky_relu(x) |
| 114 | |
| 115 | if patch == True: |
| 116 | x = slim.convolution2d(x, 1, [1, 1], activation_fn=None) |
| 117 | else: |
| 118 | x = tf.reduce_mean(x, axis=[1, 2]) |
| 119 | x = slim.fully_connected(x, 1, activation_fn=None) |
| 120 | |
| 121 | return x |
| 122 | |
| 123 | |
| 124 |
nothing calls this directly
no outgoing calls
no test coverage detected