HRNet / HigherHRNet-Human-Pose-Estimation

This is an official implementation of our CVPR 2020 paper "HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation" (https://arxiv.org/abs/1908.10357)
MIT License
1.35k stars 272 forks source link

Use valid to reason about the data, and the result looks wrong. #110

Open cena001plus opened 1 year ago

cena001plus commented 1 year ago

When using valid inference, I always get a lot of noise points, so I set the threshold DETECTION_THRESHOLD and changed it from 0.5 to 0.9, and the same graph did not get the result as expected.

When the threshold is raised, the interference is reduced, but the real results are filtered out.

DETECTION_THRESHOLD=0.5 image

DETECTION_THRESHOLD=0.9 image