how to crop images when using detection human result?
lucasjinreal opened this issue · comments
Hi, when using detection human result for E2E mesh regression on videos. How to resize the detected human box to target 224x244 input? if a human detected box is rectangle, how to resize it? How to determine the padding values on horizontal and vertical?
I think GraphCMR gives a good tutorial on how to do it. Please refer to their code: https://github.com/nkolot/GraphCMR/blob/master/demo.py