Enhanced Self-Perception in Mixed Reality: Egocentric Arm Segmentation and Database with Automatic Labeling

Abstract

In this study, we focus on the egocentric segmentation of arms to improve self-perception in Augmented Virtuality (AV). The main contributions of this work are: $i$ ) a comprehensive survey of segmentation algorithms for AV; $ii$ ) an Egocentric Arm Segmentation Dataset (EgoArm), composed of more than 10, 000 images, demographically inclusive (variations of skin color, and gender), and open for research purposes. We also provide all details required for the automated generation of groundtruth and semi-synthetic images; $iii$ ) the proposal of a deep learning network to segment arms in AV; $iv$ ) a detailed quantitative and qualitative evaluation to showcase the usefulness of the deep network and EgoArm dataset, reporting results on different real egocentric hand datasets, including GTEA Gaze+, EDSH, EgoHands, Ego Youtube Hands, THU-Read, TEgO, FPAB, and Ego Gesture, which allow for direct comparisons with existing approaches using color or depth. Results confirm the suitability of the EgoArm dataset for this task, achieving improvements up to 40% with respect to the baseline network, depending on the particular dataset. Results also suggest that, while approaches based on color or depth can work under controlled conditions (lack of occlusion, uniform lighting, only objects of interest in the near range, controlled background, etc.), deep learning is more robust in real AV applications