Thanks for the great work.
In section 4.5 of the published paper, the different datasets were combined to train a model in a supervised way. However, the class labels of the different datasets are different, i.e. COCO, VOC, and Cityscapes, so how to figure out this problem during training?
Looking forward to your reply.