Dynamic Iterative Refinement for Efficient 3D Hand Pose Estimation

John Yang, Yash Bhalgat, Simyung Chang, Fatih Porikli, Nojun Kwak (or view on SciRate)

Main idea

We propose a tiny deep network of which partial layers are recursively exploited for refining its previous estimations. During its iterative refinements, we employ learned gating criteria to decide whether to exit from the weight-sharing loop, allowing per-sample adaptation in our model. We also predict and exploit uncertainty estimations in the gating mechanism.


While hand pose estimation is a critical component of most interactive extended reality and gesture recognition systems, contemporary approaches are not optimized for computational and memory efficiency.


With the proposed setting, our method consistently outperforms state-of-the-art 2D/3D hand pose estimation approaches in terms of both accuracy and efficiency for widely used benchmarks.

tags: 3D-Computer-Vision  Hand-Pose-Estimation  Efficient-Deep-Learning