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

A question about output. #36

Closed YuQi9797 closed 3 years ago

YuQi9797 commented 3 years ago

For the validation stage, is the output of the model concatenate as the heat diagram of all stages? Or just output the heat map of the last phase?

crockwell commented 3 years ago

Only the output from the last phase. Each Hourglass uses as part of its input the previous heatmap output, so it is incorporated this way -- predictions tend to improve at later hourglasses.