YOLO Label Assist is a tool to help you create labeled images with the help of a trained model. This way you can increase your training set without having to manually mark each object. Instead you only have to fix wrong ones.
- in
config.jsonenter the process name you want to take screenshots of. - in
/trainfiles/place your .cfg, .names and .weights files of your current model, named after the process name - start YOLOLA
- record screenshots
- possible modes:
- Prediction: use the predicted boxes
- AllSelectedType: use the predicted boxes, but change the type to the selected one.
- NoObjects: save with no boxes
- open LabelImg in the
/img/folder and correct the boxes
This tool is based on this programm by Trombov