search

facebookresearch / InterHand2.6M

Official PyTorch implementation of "InterHand2.6M: A Dataset and Baseline for 3D Interacting Hand Pose Estimation from a Single RGB Image", ECCV 2020
Other
676 stars 92 forks source link

Why are the bbox of the rootnet outputs the same as the gt annotation in val set and test set? #91

Closed MengHao666 closed 2 years ago

MengHao666 commented 2 years ago

I am so confused about that. Could you tell me the reason?

mks0601 commented 2 years ago

RootNet does not predict bboxes. It predicts only root joint depths from a cropped hand image, and GT bboxes are used to prepare inputs of RootNet. The bboxes in the json files are GT bboxes, used to prepare inputs of RootNet.

MengHao666 commented 2 years ago

e json files are GT bboxes, used to prepare inputs of RootNet.

Oh, thanks for quick response. So all the bounding box used here are from GT annotations.