Is it possible to detect mask and keypoints at the same time?

Hi, I am learning the TORCHVISION OBJECT DETECTION FINETUNING TUTORIAL which is a great tutorial for mask-rcnn.

Now I want to make keypoint predictions along side the mask predictions, is it possible?

By the way, I have noticed that KeypointRCNN provided by torchvision is inherited from FasterRCNN while MaskRCNN is also inherited from FasterRCNN. So KeypointRCNN ignore the masks variable in targets.

1 Like

Any updates on this?