Domain adaptation through anatomical constraints for 3d human pose estimation under the cover

Alexander Bigalke, Lasse Hansen, Jasper Diesel, Mattias P Heinrich
Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, PMLR 172:173-187, 2022.

Abstract

Domain adaptation has the potential to overcome the expensive or even infeasible labeling of target data by transferring knowledge from a labeled source domain. In this work, we address domain adaptation in the context of point cloud-based 3D human pose estimation, whose clinical applicability is severely limited by a lack of labeled training data. Unlike the mainstream approach of domain-invariant feature learning, we propose to guide the learning process in the target domain through weak supervision, based on prior knowledge about human anatomy. We embed this prior knowledge into a novel loss function that encourages network predictions to match the statistics of an anatomically plausible skeleton. Specifically, we formulate three loss functions that penalize asymmetric limb lengths, implausible joint angles, and implausible bone lengths. We evaluate the method on a public lying pose dataset (SLP), adapting from uncovered patients in the source to covered patients in the target domain. Our method outperforms diverse state-of-the-art domain adaptation techniques and improves the baseline model by 26% while reducing the gap to a fully supervised model by 54%. Source code is available at https://github.com/multimodallearning/da-3dhpe-anatomy.

Cite this Paper

BibTeX
@InProceedings{pmlr-v172-bigalke22a, title = {Domain adaptation through anatomical constraints for 3d human pose estimation under the cover}, author = {Bigalke, Alexander and Hansen, Lasse and Diesel, Jasper and Heinrich, Mattias P}, booktitle = {Proceedings of The 5th International Conference on Medical Imaging with Deep Learning}, pages = {173--187}, year = {2022}, editor = {Konukoglu, Ender and Menze, Bjoern and Venkataraman, Archana and Baumgartner, Christian and Dou, Qi and Albarqouni, Shadi}, volume = {172}, series = {Proceedings of Machine Learning Research}, month = {06--08 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v172/bigalke22a/bigalke22a.pdf}, url = {https://proceedings.mlr.press/v172/bigalke22a.html}, abstract = {Domain adaptation has the potential to overcome the expensive or even infeasible labeling of target data by transferring knowledge from a labeled source domain. In this work, we address domain adaptation in the context of point cloud-based 3D human pose estimation, whose clinical applicability is severely limited by a lack of labeled training data. Unlike the mainstream approach of domain-invariant feature learning, we propose to guide the learning process in the target domain through weak supervision, based on prior knowledge about human anatomy. We embed this prior knowledge into a novel loss function that encourages network predictions to match the statistics of an anatomically plausible skeleton. Specifically, we formulate three loss functions that penalize asymmetric limb lengths, implausible joint angles, and implausible bone lengths. We evaluate the method on a public lying pose dataset (SLP), adapting from uncovered patients in the source to covered patients in the target domain. Our method outperforms diverse state-of-the-art domain adaptation techniques and improves the baseline model by 26% while reducing the gap to a fully supervised model by 54%. Source code is available at https://github.com/multimodallearning/da-3dhpe-anatomy.} }
Endnote
%0 Conference Paper %T Domain adaptation through anatomical constraints for 3d human pose estimation under the cover %A Alexander Bigalke %A Lasse Hansen %A Jasper Diesel %A Mattias P Heinrich %B Proceedings of The 5th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2022 %E Ender Konukoglu %E Bjoern Menze %E Archana Venkataraman %E Christian Baumgartner %E Qi Dou %E Shadi Albarqouni %F pmlr-v172-bigalke22a %I PMLR %P 173--187 %U https://proceedings.mlr.press/v172/bigalke22a.html %V 172 %X Domain adaptation has the potential to overcome the expensive or even infeasible labeling of target data by transferring knowledge from a labeled source domain. In this work, we address domain adaptation in the context of point cloud-based 3D human pose estimation, whose clinical applicability is severely limited by a lack of labeled training data. Unlike the mainstream approach of domain-invariant feature learning, we propose to guide the learning process in the target domain through weak supervision, based on prior knowledge about human anatomy. We embed this prior knowledge into a novel loss function that encourages network predictions to match the statistics of an anatomically plausible skeleton. Specifically, we formulate three loss functions that penalize asymmetric limb lengths, implausible joint angles, and implausible bone lengths. We evaluate the method on a public lying pose dataset (SLP), adapting from uncovered patients in the source to covered patients in the target domain. Our method outperforms diverse state-of-the-art domain adaptation techniques and improves the baseline model by 26% while reducing the gap to a fully supervised model by 54%. Source code is available at https://github.com/multimodallearning/da-3dhpe-anatomy.
APA
Bigalke, A., Hansen, L., Diesel, J. & Heinrich, M.P.. (2022). Domain adaptation through anatomical constraints for 3d human pose estimation under the cover. Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 172:173-187 Available from https://proceedings.mlr.press/v172/bigalke22a.html.

Related Material