2016/03/30 by Alejandro Betancourt, Natalia Díaz-Rodríguez, Betancourt, Alejandro +9 · 1 citation
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Hand Gesture Recognition Systems #Human Pose and Action Recognition #Video Surveillance and Tracking Methods
paper · pdf · doi:10.48550/arxiv.1603.09200
openalex publication_date 2016/03/30 · openalex created_date 2022/10/04 · openalex updated_date 2026/07/28
Wearable cameras stand out as one of the most promising devices for the\nupcoming years, and as a consequence, the demand of computer algorithms to\nautomatically understand the videos recorded with them is increasing quickly.\nAn automatic understanding of these videos is not an easy task, and its mobile\nnature implies important challenges to be faced, such as the changing light\nconditions and the unrestricted locations recorded. This paper proposes an\nunsupervised strategy based on global features and manifold learning to endow\nwearable cameras with contextual information regarding the light conditions and\nthe location captured. Results show that non-linear manifold methods can\ncapture contextual patterns from global features without compromising large\ncomputational resources. The proposed strategy is used, as an application case,\nas a switching mechanism to improve the hand-detection problem in egocentric\nvideos.\n