Hands Deep in Deep Learning for Hand Pose Estimation

TL;DR

This paper presents new CNN architectures for 3D hand pose estimation from depth maps, incorporating priors and context to improve accuracy and efficiency, outperforming previous methods on benchmarks.

Contribution

Introduces CNN architectures with pose priors and context modeling for improved 3D hand pose estimation from depth images.

Findings

Significant accuracy improvements over state-of-the-art methods.

Enhanced reliability in hand pose predictions.

Reduced computation times for real-time applications.

Abstract

We introduce and evaluate several architectures for Convolutional Neural Networks to predict the 3D joint locations of a hand given a depth map. We first show that a prior on the 3D pose can be easily introduced and significantly improves the accuracy and reliability of the predictions. We also show how to use context efficiently to deal with ambiguities between fingers. These two contributions allow us to significantly outperform the state-of-the-art on several challenging benchmarks, both in terms of accuracy and computation times.