Keras 使用 Lambda层详解

我就废话不多说了，大家还是直接看代码吧！

from tensorflow.python.keras.models import sequential, model
from tensorflow.python.keras.layers import dense, flatten, conv2d, maxpool2d, dropout, conv2dtranspose, lambda, input, reshape, add, multiply
from tensorflow.python.keras.optimizers import adam
 
def deconv(x):
  height = x.get_shape()[1].value
  width = x.get_shape()[2].value
  
  new_height = height*2
  new_width = width*2
  
  x_resized = tf.image.resize_images(x, [new_height, new_width], tf.image.resizemethod.nearest_neighbor)
  
  return x_resized
 
def generator(scope='generator'):
  imgs_noise = input(shape=inputs_shape)
  x = conv2d(filters=32, kernel_size=(9,9), strides=(1,1), padding='same', activation='relu')(imgs_noise)
  x = conv2d(filters=64, kernel_size=(3,3), strides=(2,2), padding='same', activation='relu')(x)
  x = conv2d(filters=128, kernel_size=(3,3), strides=(2,2), padding='same', activation='relu')(x)
 
  x1 = conv2d(filters=128, kernel_size=(3,3), strides=(1,1), padding='same', activation='relu')(x)
  x1 = conv2d(filters=128, kernel_size=(3,3), strides=(1,1), padding='same', activation='relu')(x1)
  x2 = add()([x1, x])
 
  x3 = conv2d(filters=128, kernel_size=(3,3), strides=(1,1), padding='same', activation='relu')(x2)
  x3 = conv2d(filters=128, kernel_size=(3,3), strides=(1,1), padding='same', activation='relu')(x3)
  x4 = add()([x3, x2])
 
  x5 = conv2d(filters=128, kernel_size=(3,3), strides=(1,1), padding='same', activation='relu')(x4)
  x5 = conv2d(filters=128, kernel_size=(3,3), strides=(1,1), padding='same', activation='relu')(x5)
  x6 = add()([x5, x4])
 
  x = maxpool2d(pool_size=(2,2))(x6)
 
  x = lambda(deconv)(x)
  x = conv2d(filters=64, kernel_size=(3, 3), strides=(1,1), padding='same',activation='relu')(x)
  x = lambda(deconv)(x)
  x = conv2d(filters=32, kernel_size=(3, 3), strides=(1,1), padding='same',activation='relu')(x)
  x = lambda(deconv)(x)
  x = conv2d(filters=3, kernel_size=(3, 3), strides=(1, 1), padding='same',activation='tanh')(x)
 
  x = lambda(lambda x: x+1)(x)
  y = lambda(lambda x: x*127.5)(x)
  
  model = model(inputs=imgs_noise, outputs=y)
  model.summary()
  
  return model
 
my_generator = generator()
my_generator.compile(loss='binary_crossentropy', optimizer=adam(0.7, decay=1e-3), metrics=['accuracy'])

补充知识：含有lambda自定义层keras模型，保存遇到的问题及解决方案

一，许多应用，keras含有的层已经不能满足要求，需要透过lambda自定义层来实现一些layer，这个情况下，只能保存模型的权重，无法使用model.save来保存模型。保存时会报

typeerror: can't pickle _thread.rlock objects

二，解决方案，为了便于后续的部署，可以转成tensorflow的pb进行部署。

from keras.models import load_model
import tensorflow as tf
import os, sys
from keras import backend as k
from tensorflow.python.framework import graph_util, graph_io

def h5_to_pb(h5_weight_path, output_dir, out_prefix="output_", log_tensorboard=true):
  if not os.path.exists(output_dir):
    os.mkdir(output_dir)
  h5_model = build_model()
  h5_model.load_weights(h5_weight_path)
  out_nodes = []
  for i in range(len(h5_model.outputs)):
    out_nodes.append(out_prefix + str(i + 1))
    tf.identity(h5_model.output[i], out_prefix + str(i + 1))
  model_name = os.path.splitext(os.path.split(h5_weight_path)[-1])[0] + '.pb'
  sess = k.get_session()
  init_graph = sess.graph.as_graph_def()
  main_graph = graph_util.convert_variables_to_constants(sess, init_graph, out_nodes)
  graph_io.write_graph(main_graph, output_dir, name=model_name, as_text=false)
  if log_tensorboard:
    from tensorflow.python.tools import import_pb_to_tensorboard
    import_pb_to_tensorboard.import_to_tensorboard(os.path.join(output_dir, model_name), output_dir)

def build_model():
  inputs = input(shape=(784,), name='input_img')
  x = dense(64, activation='relu')(inputs)
  x = dense(64, activation='relu')(x)
  y = dense(10, activation='softmax')(x)
  h5_model = model(inputs=inputs, outputs=y)
  return h5_model

if __name__ == '__main__':
  if len(sys.argv) == 3:
    # usage: python3 h5_to_pb.py h5_weight_path output_dir
    h5_to_pb(h5_weight_path=sys.argv[1], output_dir=sys.argv[2])

以上这篇keras 使用 lambda层详解就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持。