TensorFlow实现FCN

最新推荐文章于 2025-06-30 11:25:41 发布

蓬莱道人

最新推荐文章于 2025-06-30 11:25:41 发布

阅读量2w

点赞数 18

CC 4.0 BY-SA版权

分类专栏： Semantic Segmentation

本文链接：https://blog.csdn.net/MOU_IT/article/details/81073149

FCN的网络结构：

FCN全名叫做全卷机神经网络，它在经典的VGGNet的基础上，把VGG网络最后的全连接层全部去掉，换为卷积层。为了能对图像进行分割，FCN对卷积后的结果进行了反卷积，生成和原图一样的尺寸输出，然后经过softmax就能对每个像素进行分类。具体的网络结果如下：

论文参考《Fully Convolutional Networks for Semantic Segmentation》，代码实现参考：https://github.com/shekkizh/FCN.tensorflow

代码详解：

代码的实现有四个python文件，分别是FCN.py、BatchDatasetReader.py、TensorFlowUtils.py、read_MITSceneParsingData.py。将这四个文件放在一个当前目录 . 下，然后去这里下载VGG网络的权重参数，下载好后的文件路径为./Model_zoo/imagenet-vgg-verydeep-19.mat，然后去这里下载训练会用到的数据集，并解压到路径: ./Data_zoo/MIT_SceneParsing/ADEChallengeData2016。训练时把FCN.py中的全局变量mode该为“train”，运行该文件。测试时改为“visualize”运行即可。

FCN.py为主文件，代码如下：

from __future__ import print_function
import tensorflow as tf
import numpy as np

import TensorflowUtils as utils
import read_MITSceneParsingData as scene_parsing
import datetime
import BatchDatsetReader as dataset
from six.moves import xrange

batch_size=2                             # batch 大小
logs_dir="logs/"
data_dir= "Data_zoo/MIT_SceneParsing/"   # 存放数据集的路径，需要提前下载
data_name="ADEChallengeData2016"
learning_rate=1e-4                                           # 学习率
model_path="Model_zoo/imagenet-vgg-verydeep-19.mat"          # VGG网络参数文件，需要提前下载
debug= False
mode='train'                             # 训练模式train | visualize

MODEL_URL = 'http://www.vlfeat.org/matconvnet/models/beta16/imagenet-vgg-verydeep-19.mat' #训练好的VGGNet参数

MAX_ITERATION = int(1e5 + 1)   # 最大迭代次数
NUM_OF_CLASSESS = 151          # 类的个数
IMAGE_SIZE = 224               # 图像尺寸

# 根据载入的权重建立原始的 VGGNet 的网络
def vgg_net(weights, image):
    layers = (
        'conv1_1', 'relu1_1', 'conv1_2', 'relu1_2', 'pool1',
        'conv2_1', 'relu2_1', 'conv2_2', 'relu2_2', 'pool2',
        'conv3_1', 'relu3_1', 'conv3_2', 'relu3_2', 'conv3_3','relu3_3', 'conv3_4', 'relu3_4', 'pool3',
        'conv4_1', 'relu4_1', 'conv4_2', 'relu4_2', 'conv4_3','relu4_3', 'conv4_4', 'relu4_4', 'pool4',
        'conv5_1', 'relu5_1', 'conv5_2', 'relu5_2', 'conv5_3','relu5_3', 'conv5_4', 'relu5_4'
    )

    net = {}
    current = image
    for i, name in enumerate(layers):
        kind = name[:4]
        if kind == 'conv':
            kernels, bias = weights[i][0][0][0][0]
            # matconvnet: weights are [width, height, in_channels, out_channels]
            # tensorflow: weights are [height, width, in_channels, out_channels]
            kernels = utils.get_variable(np.transpose(kernels, (1, 0, 2, 3)), name=name + "_w")
            bias = utils.get_variable(bias.reshape(-1), name=name + "_b")
            current = utils.conv2d_basic(current, kernels, bias)
            print ("当前形状：",np.shape(current))
        elif kind == 'relu':
            current = tf.nn.relu(current, name=name)
            if debug:
                utils.add_activation_summary(current)
        elif kind == 'pool':
            current = utils.avg_pool_2x2(current)
            print ("当前形状：",np.shape(current))
        net[name] = current

    return net


# FCN的网络结构定义，网络中用到的参数是迁移VGG训练好的参数
def inference(image, keep_prob):
    """
    Semantic segmentation network definition
    :param image: input image. Should have values in range 0-255
    :param keep_prob:
    :return:
    """
    # 加载模型数据    
    print ("原始图像：",np.shape(image))
    model_data = utils.get_model_data(model_path)

    mean = model_data['normalization'][0][0][0]
    mean_pixel = np.mean(mean, axis=(0, 1))

    weights = np.squeeze(model_data['layers'])

    # 图像预处理
    processed_image = utils.process_image(image, mean_pixel)
    print ("预处理后的图像:",np.shape(processed_image))

    with tf.variable_scope("inference"):
        # 建立原始的VGGNet-19网络

        print ("开始建立VGG网络：")
        image_net = vgg_net(weights, processed_image)

        # 在VGGNet-19之后添加 一个池化层和三个卷积层
        conv_final_layer = image_net["conv5_3"]
        print ("VGG处理后的图像：",np.shape(conv_final_layer))

        pool5 = utils.max_pool_2x2(conv_final_layer)

        print ("pool5：",np.shape(pool5))

        W6 = utils.weight_variable([7, 7, 512, 4096], name="W6")
        b6 = utils.bias_variable([4096], name="b6")
        conv6 = utils.conv2d_basic(pool5, W6, b6)
        relu6 = tf.nn.relu(conv6, name="relu6")
        if debug:
            utils.add_activation_summary(relu6)
        relu_dropout6 = tf.nn.dropout(relu6, keep_prob=keep_prob)

        print ("conv6:",np.shape(relu_dropout6))

        W7 = utils.weight_variable([1, 1