Pose-Oriented Transformer with Uncertainty-Guided Refinement for 2D-to-3D Human Pose Estimation
Han Li, Bowen Shi, Wenrui Dai, Hongwei Zheng, Botao Wang, Yu Sun, Min Guo, Chenglin Li
Abstract
There has been a recent surge of interest in introducing transformers to 3D human pose estimation (HPE) due to their powerful capabilities in modeling long-term dependencies. However, existing transformer-based methods treat body joints as equally important inputs and ignore the prior knowledge of human skeleton topology in the self-attention mechanism. To tackle this issue, in this paper, we propose a Pose-Oriented Transformer (POT) with uncertainty guided refinement for 3D HPE. Specifically, we first develop novel pose-oriented self-attention mechanism and distance-related position embedding for POT to explicitly exploit the human skeleton topology. The pose-oriented self-attention mechanism explicitly models the topological interactions between body joints, whereas the distance-related position embedding encodes the distance of joints to the root joint to distinguish groups of joints with different difficulties in regression. Furthermore, we present an Uncertainty-Guided Refinement Network (UGRN) to refine pose predictions from POT, especially for the difficult joints, by considering the estimated uncertainty of each joint with uncertainty-guided sampling strategy and self-attention mechanism. Extensive experiments demonstrate that our method significantly outperforms the state-of-the-art methods with reduced model parameters on 3D HPE benchmarks such as Human3.6M and MPI-INF-3DHP.
BibTeX
@article{Li_Shi_Dai_Zheng_Wang_Sun_Guo_Li_Zou_Xiong_2023, title={Pose-Oriented Transformer with Uncertainty-Guided Refinement for 2D-to-3D Human Pose Estimation}, volume={37}, url={https://ojs.aaai.org/index.php/AAAI/article/view/25213}, DOI={10.1609/aaai.v37i1.25213}, abstractNote={There has been a recent surge of interest in introducing transformers to 3D human pose estimation (HPE) due to their powerful capabilities in modeling long-term dependencies. However, existing transformer-based methods treat body joints as equally important inputs and ignore the prior knowledge of human skeleton topology in the self-attention mechanism. To tackle this issue, in this paper, we propose a Pose-Oriented Transformer (POT) with uncertainty guided refinement for 3D HPE. Specifically, we first develop novel pose-oriented self-attention mechanism and distance-related position embedding for POT to explicitly exploit the human skeleton topology. The pose-oriented self-attention mechanism explicitly models the topological interactions between body joints, whereas the distance-related position embedding encodes the distance of joints to the root joint to distinguish groups of joints with different difficulties in regression. Furthermore, we present an Uncertainty-Guided Refinement Network (UGRN) to refine pose predictions from POT, especially for the difficult joints, by considering the estimated uncertainty of each joint with uncertainty-guided sampling strategy and self-attention mechanism. Extensive experiments demonstrate that our method significantly outperforms the state-of-the-art methods with reduced model parameters on 3D HPE benchmarks such as Human3.6M and MPI-INF-3DHP.}, number={1}, journal={Proceedings of the AAAI Conference on Artificial Intelligence}, author={Li, Han and Shi, Bowen and Dai, Wenrui and Zheng, Hongwei and Wang, Botao and Sun, Yu and Guo, Min and Li, Chenglin and Zou, Junni and Xiong, Hongkai}, year={2023}, month={Jun.}, pages={1296-1304} }