Zhou Zhishan ; zhou Shihao. ; Lv Zhi ; Zou Minqiang ; Tang Yao ; Liang Jiajun

Abstract

3D hand pose estimation has found broad application in areas such as gesturerecognition and human-machine interaction tasks. As performance improves, thecomplexity of the systems also increases, which can limit the comparativeanalysis and practical implementation of these methods. In this paper, wepropose a simple yet effective baseline that not only surpassesstate-of-the-art (SOTA) methods but also demonstrates computational efficiency.To establish this baseline, we abstract existing work into two components: atoken generator and a mesh regressor, and then examine their core structures. Acore structure, in this context, is one that fulfills intrinsic functions,brings about significant improvements, and achieves excellent performancewithout unnecessary complexities. Our proposed approach is decoupled from anymodifications to the backbone, making it adaptable to any modern models. Ourmethod outperforms existing solutions, achieving state-of-the-art (SOTA)results across multiple datasets. On the FreiHAND dataset, our approachproduced a PA-MPJPE of 5.7mm and a PA-MPVPE of 6.0mm. Similarly, on the Dexycbdataset, we observed a PA-MPJPE of 5.5mm and a PA-MPVPE of 5.0mm. As forperformance speed, our method reached up to 33 frames per second (fps) whenusing HRNet and up to 70 fps when employing FastViT-MA36