Convert Caffe models to Keras models

readme.md

##Convert Caffe VGG and other models to Keras Use this keras fork to convert the model.

How to convert the VGG16 model:

python -m keras.caffe.caffe2keras -load_path [model_directory] -prototxt [protoxt_name] -caffemodel [caffemodel_name] -store_path [output_path] -network_type Sequential

Load the keras model file using loadvgg.py

loadvgg.py

#How to load the model
def build_model(img_width=224, img_height=224):
  from keras.models import Sequential
  from keras.layers import Convolution2D, ZeroPadding2D, MaxPooling2D, Activation
  
  model = Sequential()
  model = Sequential()
  model.add(ZeroPadding2D((1,1),input_shape=(3,img_width,img_height)))
  model.add(Convolution2D(64, 3, 3, activation='relu', name='conv1_1'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(64, 3, 3, activation='relu', name='conv1_2'))
  model.add(Activation('relu'))
  model.add(MaxPooling2D((2, 2), strides=(2, 2)))
  
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(128, 3, 3, activation='relu', name='conv2_1'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(128, 3, 3, activation='relu', name='conv2_2'))
  model.add(Activation('relu'))
  model.add(MaxPooling2D((2, 2), strides=(2, 2)))
  
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_1'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_2'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(256, 3, 3, activation='relu', name='conv3_3'))
  model.add(Activation('relu'))
  model.add(MaxPooling2D((2, 2), strides=(2, 2)))
  
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_1'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_2'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(512, 3, 3, activation='relu', name='conv4_3'))
  model.add(Activation('relu'))
  model.add(MaxPooling2D((2, 2), strides=(2, 2)))
  
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_1'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_2'))
  model.add(Activation('relu'))
  model.add(ZeroPadding2D((1, 1)))
  model.add(Convolution2D(512, 3, 3, activation='relu', name='conv5_3'))
  model.add(Activation('relu'))
  model.add(MaxPooling2D((2, 2), strides=(2, 2)))
  model.add(Flatten())
  model.add(Dense(4096))
  model.add(Activation('relu'))
  model.add(Dropout(0.5))
  model.add(Dense(4096)
  model.add(Activation('relu'))
  model.add(Dropout(0.5))
  model.add(Dense(1000)
  model.add(Activation('softmax'))
  return model

def loadweights(model, weights_path):
  f = h5py.File(weights_path)
  for k in range(f.attrs['nb_layers']):
    if k >= len(model.layers):
      # we don't look at the last (fully-connected) layers in the savefile
      break
    g = f['layer_{}'.format(k)]
    weights = [g['param_{}'.format(p)] for p in range(g.attrs['nb_params'])]
    try:
      model.layers[k].set_weights(weights)
    except Exception:
      pass
  f.close()

  
model = build_model()
weights_path = '/path/to/your/keras/model.h5'
loadweights(model, weights_path)

Nice Job, but why the activation functions of Convolution2D layers are duplicated? The activation of each Convolution2D layer is defined both in 'activation=' argument of the layer as well as 'Activation' layer after that.