Why the multi-view errors in Table 2 are even lower than the best hypothesis errors in Table 1?

chaneyddtt / Generating-Multiple-Hypotheses-for-3D-Human-Pose-Estimation-with-Mixture-Density-Network

Code for our CVPR2019 paper: Generating Multiple Hypotheses for 3D Human Pose Estimation with Mixture Density Network

MIT License

97 stars 9 forks source link

Why the multi-view errors in Table 2 are even lower than the best hypothesis errors in Table 1? #7

Open luzzou opened 4 years ago

luzzou commented 4 years ago

Screenshot from 2019-12-23 14-26-19

chaneyddtt commented 4 years ago

Hi @trumDog, We first select the hypotheses under different viewpoints by using the consistent constraint. Then we take the average value of the selected hypotheses under different viewpoints for the same 3d pose. The results is better than the best hypothesis mainly because of the averaging process.