search

princeton-vl / pytorch_stacked_hourglass

Pytorch implementation of the ECCV 2016 paper "Stacked Hourglass Networks for Human Pose Estimation"
BSD 3-Clause "New" or "Revised" License
469 stars 94 forks source link

Why do we need to predict twice？ #46

Closed zhoujinhai closed 1 year ago

zhoujinhai commented 1 year ago

In test.py, I find forward twice and get tmp1, tmp2. I want to know why you're doing this？ image

crockwell commented 1 year ago

Hey - good catch! So actually running the model on both the image and the flipped image, and averaging the keypoints ends up doing better than just running once. So, it is run twice.

zhoujinhai commented 1 year ago

Thanks reploy.