Instead of using the given bounding_box_2d corner coordinates, how would one go about writing the screen space position of a prim/Xform? For my use case, I am randomizing object positions and have Xforms attached around them and need to write the Xform’s position in screen space, even if they are occluded.
Hello, can you please elaborate what do you mean with the screen-space position of a prim/XForm?
Hi! I suppose what I am looking for is some way to have a keypoint labeler, similar to what is used in human pose estimation. So that one can designate points in 3D space and have the coordinate of those points in the relationship to the exported image. Here is an implementation in the Unity Perception package.
I don’t see in the documentation a direct implementation of this so I was wondering if one could develop a keypoint labeler similar to the way 2D bounding boxes are implemented, where the corners of the boxes are exported in relationship to the written image export.
1 Like
Hello, this discussion may be helpful as we go over the boundaries of the 3D bounding box. Insight into world to camera transform for 3D bounding box
We don’t have by default the keypoints but you may modify a writer to give you the keypoints.
1 Like