search

xingyizhou / pytorch-pose-hg-3d

PyTorch implementation for 3D human pose estimation
GNU General Public License v3.0
614 stars 143 forks source link

I am confused about the principle where you transform the images to cropped image #58

Closed hutao568 closed 5 years ago

hutao568 commented 5 years ago

hello ,zhou.I am confused about the principle where you transform the images to cropped image in your {Project}/src/lib/dataset/mpii.pi file.I know if you get three points' coordinate before and after transform,you can get the affine matrix.But i don't understand how you uses center,scale,rota to get three points' coordinate before and after transform?Could you tell me the principle or you can tell me where i can find the answer?

hutao568 commented 5 years ago

I understand the place, you take the center of the original image and the center of the upper boundary and the third generated point