The hardware environment of this paper is DELL, GPU 1060TI, and the software environment iswindows10 64-bitOperating system, python 3.

First install the labelme based on anaconda in the windows system:

# python3
conda create --name=labelme python=3.6

source activate labelme

pip install pyqt5  

pip install labelm

Enter in anaconda promote after installation is completelabelmeYou can open the graphical user interface as shown below:

At this point, open the image that needs to be labeled and mark the image.

After the image annotation is completed, it is saved as a json format file, but there is no way to directly use the json format file, so the json file needs to be processed later:

1. Find the location of the json file and set the location of the processed file

Labelme_json_to_dataset storage location / file name .json -o processed file location / file name

After processing, 5 files will appear in the folder, and then 5 files will be processed.

The five files are: original file .png, info.yaml, label.png, label_names.txt, label_viz.png

Where label.png is the label image of the image

2. Coloring label.png

Batch processing using the python program, the program is as follows:

import PIL.Image
import numpy as np
from skimage import io,data,color
import matplotlib.pyplot as plt

for i in range(1,7):
         Img=PIL.Image.open("file path/label.png"%i)
    img=np.array(img)
         Dst=color.label2rgb(img,bg_label=0,bg_color=(0,0,0))#The background label is 0 and the color is black.
         Io.imsave("file path.png"%i,dst)

The result is as follows:

The original image is too large to be uploaded, and the original image is an aerial image of the aerial image.

3. Using Matlab to convert 24-bit map to 8-bit map

Dirs=dir('file path/*.JPG');
for n=1:numel(dirs)
         Strname=strcat('file path/',dirs(n).name);
    img=imread(strname);
    [x,map]=rgb2ind(img,256);
         Newname=strcat('result storage path/',dirs(n).name);
    imwrite(x,map,newname,'JPG');
end

The converted image is subjected to training such as rotation division, and the like.