Observing and Understanding Hands in Action

The fifth edition of this ICCV 2019 workshop aims at gathering researchers who work on 2D/3D hand detection, segmentation, pose estimation, and tracking problems and its applications. This edition will emphasize hand-object interaction and RGB-based pose estimation, via invited speakers, accepted papers, and a challenge to be organised. Development of RGB-D sensors and camera miniaturization (wearable cameras, smart phones, ubiquitous computing) have opened the door to a whole new range of technologies and applications which require detecting hands and recognizing hand poses in a variety of scenarios, including AR/VR, assistive car driving, robot grasping, and health care. A majority of hand tracking data sets and papers have been focused on near-range front-on scenarios, where a single hand or multiple hands appear visible or under moderate occlusions. Most existing methods fail to address severe occlusions under hand-object or hand-hand interaction scenarios. In parallel, RGB-based (cf. depth-based) hand pose estimation has been increasingly important in recent literature. New benchmarks and methods have been proposed for the RGB-modality, yet there remain many challenges. Our goal is to push the boundaries of 3D hand pose estimation and its relevant problems under hand-object interaction scenarios using depth images and or RGB images. We cover up a "breadth of applications" including sign language recognition, desktop interaction, egocentric views, object manipulations, far range and over-the-shoulder driver footage. The relevant topics include: