models.py 文件源码

def DenseNet_FCN(input_shape=None, weight_decay=1E-4,
                 batch_momentum=0.9, batch_shape=None, classes=21,
                 include_top=False, activation='sigmoid'):
    if include_top is True:
        # TODO(ahundt) Softmax is pre-applied, so need different train, inference, evaluate.
        # TODO(ahundt) for multi-label try per class sigmoid top as follows:
        # x = Reshape((row * col * classes))(x)
        # x = Activation('sigmoid')(x)
        # x = Reshape((row, col, classes))(x)
        return densenet.DenseNetFCN(input_shape=input_shape,
                                    weights=None, classes=classes,
                                    nb_layers_per_block=[4, 5, 7, 10, 12, 15],
                                    growth_rate=16,
                                    dropout_rate=0.2)

    # if batch_shape:
    #     img_input = Input(batch_shape=batch_shape)
    #     image_size = batch_shape[1:3]
    # else:
    #     img_input = Input(shape=input_shape)
    #     image_size = input_shape[0:2]

    input_shape = _obtain_input_shape(input_shape,
                                      default_size=32,
                                      min_size=16,
                                      data_format=K.image_data_format(),
                                      include_top=False)
    img_input = Input(shape=input_shape)

    x = densenet.__create_fcn_dense_net(classes, img_input,
                                        input_shape=input_shape,
                                        nb_layers_per_block=[4, 5, 7, 10, 12, 15],
                                        growth_rate=16,
                                        dropout_rate=0.2,
                                        include_top=include_top)

    x = top(x, input_shape, classes, activation, weight_decay)
    # TODO(ahundt) add weight loading
    model = Model(img_input, x, name='DenseNet_FCN')
    return model