img_layers.py

from paddle.trainer_config_helpers import *

settings(
    learning_rate=1e-3,
    batch_size=1000
)

img = data_layer(name='image', size=256*256)

img_conv = img_conv_layer(input=img, num_channels=1, num_filters=64,
                          filter_size=(32, 64), padding=(1, 0), stride=(1, 1),
                          act=LinearActivation())
img_bn = batch_norm_layer(input=img_conv, act=ReluActivation())

img_norm = img_cmrnorm_layer(input=img_bn, size=32)

img_pool = img_pool_layer(input=img_conv, pool_size=32, pool_type=MaxPooling())


outputs(img_pool, img_norm)