Open baeckgoo opened 3 weeks ago
@ehedlin Can you look into this?
Sorry for the poor code quality. These were precomputed from either SPIN or MaskRCNN. I will try to go through each item so hopefully it can be of some use.
orient
, betas
, pose
, gt_j3d
, and gt_j2d
are output by SPIN. They pretty much represent the orientation of the person relative to the camera and the shape and pose, and corresponding 3d/2d projected joint locations of the person respectively. Also I think spin_image
is some processed version of the input image but I don't remember how I got it exactly. intrinsics
is from SPIN as well and IIRC, it doesn't actually estimate the intrinsics and is used to find the 2d crop of the person.
estimated_translation
, bboxes
, and mask_rcnn
I'm less sure of but I think are output by maskrcnn where mask_rcnn
is binary occupancy of a person.
It seems pixel_annotations
is not used which is good because I dont remember what it was anyway.
I hope this is of some help and sorry once again for the code quality.
Thank you for the reply! It was helpful for me to understand the code.
Thank you for your work!
I'd like to use this code to refine the joint regressor. I need the preprocessed data files in here How can I find the files?
Thanks.