Labelme is generally marked with irregular shapes, and it can also be marked with rectangles, but sometimes there are many files in json format on hand, what should I do? Writing scripts can solve the problem, but it may be a bit troublesome for beginners, especially for children's shoes who are just learning python, and may not be able to write it for a day. In order to solve this problem, we specially developed a one-click conversion of json format to standard VOC data set format xml file on windows (software needs to be authorized to use, pay attention to WeChat public account: future independent research center, you can contact us to obtain software), and Automatically save the json file picture. After testing, the json file marked by labelme can be successfully opened in labelImg after being converted into an xml file. First we look at the software interface:
How to use: select your own json file directory and click to start conversion. After conversion, JPEGImages and Annotations folders will be automatically generated. This is the labeled VOC format data we need, so that we can get frameworks such as yolov3/yolov4/yolov5/ssd/faster-rcnn as training data sets.