deeplab implements its own remote sensing geological segmentation dataset

1. 数据集下载

下载链接
提取码：2022
Samples are stored in the datasets folderimage和label,You can view their channel and other information,Convenient to work with your own datasets.

2. 数据集预处理

参考另一篇文章：图像预处理

1. 数据集image
   1. 是24位深度
   2. 3通道RGB图像
2. 数据集label
   1. 是8位深度
   2. The value of the label is0,1,2,…(有nclass valuen-1)

If your dataset labels take the value 255,需要转化为0,1,2,等（视自己情况而定）
例如：

Background [0,0,0]--------------0

Person [192,128,128]--------------1

Bike [0,128,0]----------------------2

Car [128,128,128]----------------- 3

Drone [128,0,0]--------------------4

Boat [0,0,128]--------------------- 5

Animal [192,0,128]---------------- 6

Obstacle [192,0,0]------------------7

Construction [192,128,0]-----------8

Vegetation [0,64,0]-----------------9

Road [128,128,0]-------------------10

Sky [0,128,128]---------------------11

如果显存不够,Images can be cropped as 256大小,对应代码如下：
The cropped image is saved in image和label中

import os
import numpy as np
import cv2

images_path = './JPEGImages/'
labels_path = './SegmentationClass/'

image_files = os.listdir(images_path)

for s in image_files:
    image_path = images_path + s
    label_path = labels_path + s[:-4]+'.png'
    image = cv2.imread(image_path)
    label = cv2.imread(label_path)
    #print(image.shape)
    index = 0
    for i in range(4):
        for j in range(4):
            #print(i*256, ((i+1)*256-1), j*256, ((j+1)*256-1))
            new_image = image[i*256 : ((i+1)*256), j*256 : ((j+1)*256), :]
            new_label = label[i*256 : ((i+1)*256), j*256 : ((j+1)*256), :]
            cv2.imwrite('./image/'+ 'b_' + s[:-4] +'_'+ str(index) + '.png', new_image)
            new_label = cv2.cvtColor(new_label, cv2.COLOR_BGR2GRAY)
            cv2.imwrite('./label/'+ 'b_' + s[:-4] +'_'+ str(index) + '.png', new_label)
            index+=1
    print(s)

3. 模型下载

下载链接

4. 检查数据

1. 数据准备完成之后,首先运行voc_annotation.py文件,生成数据列表,运行命令：python voc_annotation.py,Usually no error is reported here,If an error is reported, you can check as follows：

   1. Check if the label is8位深度,如果不是,运行下面的代码：

      import os
      
      import cv2
      
      file_names = os.listdir('./SegmentationClass/')
      
      for s in file_names:
          image_path = os.path.join('./SegmentationClass/', s)
          image = cv2.imread(image_path)
          image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
          cv2.imwrite('./SegmentationClass/' + s , image)
      
      

   2. Check if there are other values ​​in the label,比如说类别为3,Then the value in the tag can only be0,1,2,如果出现其他值,就可能会报错;Even if there is no error here,There may also be problems during training,So if there is this problem,一定要记得修改.If this value is not particularly large,可以直接设为0;如果特别多,Then you need to check which step is causing the problem,If you can't find it, suggest changing the dataset.

5. 训练模型

1. 修改train.py代码
   图像的尺寸（重要）
   
   图像类别（重要）
   
   batch_sizeand other parameters as the case may be
   
2. 运行python train.py来训练
   

6. 模型评估

1. 修改deeplab.py内容,are the trained models.类别数（Including background）.图像尺寸
   
2. 修改get_miou.py内容
   
3. 运行python get_miou.py命令
   

7. 模型预测

1. 修改deeplab.py内容,are the trained models.类别数（Including background）.图像尺寸
   
2. 修改predict.py内容.name_classesIndicates the corresponding name for each category
   
3. 运行python predict.py命令