This is pytorch implemtation for maskrcnn predict keypoints and corresponding depths
run python main.py
run python kp3d_predict.py
##the modification of maskrcnn
-
the model function is keypoints_rcnn_3d.py ,in which we add a depth branch
-
mainly focus on the roi_heads.py ,in which we add depths loss ,similar to the keypoints loss
##the another way to add depth
-
the model function is keypoints_rcnn_3d2.py ,in which we add a depth branch,but share the roi_pool and head with keypoints
-
the roi_heads2.py ,in which we add a depth loss ,but share the keypoint_feature