In the yolov7-segmentation model, there are 80 classes provided by COCO, and with the yolov7-seg.pt file obtained from your GitHub account, I am able to detect and perform segmentation on these 80 classes as desired. What I want is to add my own custom class to these 80 classes, so that the yolov7-seg.pt file would have a total of 81 classes. However, I don't want to retrain the other 80 classes, I only want to train my custom class.
For example, I have labeled images and their corresponding annotations for object "A", separated into train and val sets. I only want to train this object "A" and have the yolov7-seg.pt file contain 80 classes + class "A" (81 classes in total).
This topic has been discussed in relation to YOLOv5 in previous years, but I couldn't find a clear answer or information on it.
I tried the following, but I couldn't succeed:
1- https://community.ultralytics.com/t/how-to-combine-weights-to-detect-from-multiple-datasets/38
