BigHand2.2M Benchmark: Hand Pose Dataset and State of the Art Analysis

TL;DR

This paper introduces BigHand2.2M, a large-scale, richly annotated hand pose dataset captured with a novel sensor-based method, enabling improved hand pose estimation and benchmarking across diverse scenarios.

Contribution

The paper presents a new large-scale hand pose dataset using a novel capture method with magnetic sensors, covering a wider range of natural hand poses than existing datasets.

Findings

Significant improvement in cross-benchmark hand pose estimation performance.

Enhanced egocentric hand pose estimation with CNN trained on the new dataset.

The dataset exhibits a wider and denser range of hand poses.

Abstract

In this paper we introduce a large-scale hand pose dataset, collected using a novel capture method. Existing datasets are either generated synthetically or captured using depth sensors: synthetic datasets exhibit a certain level of appearance difference from real depth images, and real datasets are limited in quantity and coverage, mainly due to the difficulty to annotate them. We propose a tracking system with six 6D magnetic sensors and inverse kinematics to automatically obtain 21-joints hand pose annotations of depth maps captured with minimal restriction on the range of motion. The capture protocol aims to fully cover the natural hand pose space. As shown in embedding plots, the new dataset exhibits a significantly wider and denser range of hand poses compared to existing benchmarks. Current state-of-the-art methods are evaluated on the dataset, and we demonstrate significant improvements in cross-benchmark performance. We also show significant improvements in egocentric hand pose estimation with a CNN trained on the new dataset.