Using only 2D landmark labels, a student pose network trained with a soft NRSfM-derived loss achieves lower depth error than its NRSfM teacher and than prior weakly supervised methods.
3d human pose es- timation = 2d pose estimation + matching
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Distill Knowledge from NRSfM for Weakly Supervised 3D Pose Learning
Using only 2D landmark labels, a student pose network trained with a soft NRSfM-derived loss achieves lower depth error than its NRSfM teacher and than prior weakly supervised methods.