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 bounding boxes of rootnet outputs same with that of annotations in machine_train branch. #94

Closed MengHao666 closed 2 years ago

MengHao666 commented 2 years ago

I found the bbox here is ame with the bbox here. Could you give some explainantions?

MengHao666 commented 2 years ago

Could anyone here answer my question?

mks0601 commented 2 years ago

RootNet predicts absolute 3D wrist location in the camera-centered coordinate system from a cropped hand image. To crop the hand image, GT bbox is used. That is why both have the same GT bboxes.

MengHao666 commented 2 years ago

RootNet predicts absolute 3D wrist location in the camera-centered coordinate system from a cropped hand image. To crop the hand image, GT bbox is used. That is why both have the same GT bboxes.

Got it! Thanks for kind reply!