I work on a semantic segmentation project where I have 3 classes. I annotated the images into the ground truth masks with labels for each class and converted t
3 classes