完整工程，deeplab v3+(tensorflow)代码全理解及其运行过程，长期更新

前提：ubuntu+tensorflow-gpu+python3.6

各种环境提前配好

1.下载工程源码

网址：https://github.com/tensorflow/models

下载时会遇到速度过慢或中间因为网络错误停止，可以换移动网络或者用迅雷下载。

2.测试环境

先添加slim路径，每次打开terminal都要加载路径

# From tensorflow/models/research/
export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim

运行测试1

# From tensorflow/models/research/
python deeplab/model_test.py

测试2

# From tensorflow/models/research/deeplab
sh local_test.sh

3.处理数据标签

处理标签为单通道，运行下面代码即可

import numpy as np
from PIL import Image
from keras.preprocessing.image import load_img, img_to_array
import os

classes = ['background', 'aeroplane', 'bicycle', 'bird', 'boat',
           'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'dining table',
           'dog', 'horse', 'motorbike', 'person', 'potted plant',
           'sheep', 'sofa', 'train', 'tv/monitor']

colormap = [[, , ], [, , ], [, , ], [, , ], [, , ],
            [, , ], [, , ], [, , ], [, , ], [, , ],
            [, , ], [, , ], [, , ], [, , ],
            [, , ], [, , ], [, , ], [, , ],
            [, , ], [, , ], [, , ]]

# 利用下面的代码，将标注的图片转换为单通道的label图像
cm2lbl = np.zeros(**)
for i, cm in enumerate(colormap):
    cm2lbl[(cm[]*+cm[])*+cm[]] = i

def image2label(im):
    # 输入为标记图像的矩阵，输出为单通道映射的label图像
    data = im.astype('int32')
    idx = (data[:, :, ]*+data[:, :, ])*+data[:, :, ]
    return np.array(cm2lbl[idx])

def change_label(label_url, label_name):

    label_img = load_img(label_url)
    label_img = img_to_array(label_img)
    label_img = image2label(label_img)  # 将图片映射为单通道数据
    print(np.max(label_img))

    label_single = Image.fromarray(label_img)
    label_single = label_single.convert('L')

    save_path = './datasets/VOC2012/Label'
    save_path = os.path.join(save_path, label_name)  # 确定保存路径及名称
    label_single.save(save_path)

val_file_path = './datasets/VOC2012/ImageSets/trainval.txt'  # 文件名存放路径
label_file_path = './datasets/VOC2012/SegmentationClass'  # 原label存放路径

with open(val_file_path, 'r') as f:
    file_names = f.readlines()
    count =
    for name in file_names:
        count +=
        name = name.strip('\n')  # 去掉换行符
        label_name = name + '.png'  # label文件名
        label_url = os.path.join(label_file_path, label_name)
        print('这是第 %s 张' % count)
        print(label_url)
        change_label(label_url, label_name)

4.运行build_voc2012_data.py 生成 .tfrecord数据

5. 运行train.py训练模型