Some critical questions about the code process of human2sim2robot code #9
Description
Hey, Tylerlum：
Thank you for your excellent contribution! We are now referring your code for our project. However, i come across three issues in your code which are very weird.
1.In your hand reconstruction and icp process(hamer_depth/run.py/run_utils.py, i noticed that you first use trimesh.ray.ray_triangle.RayMeshIntersector to get the visible hamer points and project the visible hamer points to 2d to refine the depth points. I try to adapt this function(def refine_3d_pts_with_depth) to my project, but the result is terrible, the projected 2d pixels may cover the depth points which are not belong to the rgbd hand pc. So, i want to know the meaning of this function(def refine_3d_pts_with_depth)
2.In the function of (def get_initial_transformation_estimate) in your hamer icp process, i noticed that you try to filter out the hamer points, which is what ican't understand. In my opinion, the visible hamer pc is very clear. Why should you do noise filter and calculate the biggest connected part?
The questions are just my own opinions if anything wrong. I'm looking forward to your reply!