Study Notes 1——Making Datasets
1. Data labeling tool labelimg
installation process cmd(win+R)→pip install labelimg, the specific operation is as shown in the figure below:
win+r first open the cmd window 
and 
install it directly in the window with pip install labelimg Finally, open the file to be labeled, and execute the command line labelimg predefined_classes.txt. The content of the txt file is classes. When there are many categories, you can have it, or you can directly labelimg. This is a folder based on the VOC format, and the Annotations folder is 
placed Labeled xml files, ImageSets contains the txt files of the divided training set and test set, and the images to be labeled are placed under the JPEGImages folder. After the labelimg is opened, it looks like this 2. The use of labelimg After opening the labelimg, Figure 3 
will 
pop
up Click the Annotations folder to save the annotated xml file to this folder.
Here you can change the format of the label, the figure is in xml format, click to change to yolo, json format
Click view, Auto Save mode: when switching to the next picture, the label will be saved automatically; DisplayLabels: display the label box and label; Advance mode: the marked cross will be suspended on the window Use of shortcut keys: A: switch to the 
previous
one
D: Switch to the next picture ;
del: Delete the label box
3. Start labeling
Click on the cross to mark the object. 
Start labeling 
If the marked object is difficult to recognize, you can tick difficult
Annotated xml file