Assisting group activity analysis through hand detection and identification in multiple ...
summary
A multiple egocentric video browser surfaces frames labeled by hand-mask detection via ELAN network into parallel timelines and three concrete hand-based features: individual action, joint action, and implicit (observer) action. A prototype highlights owner-only, both-owner-and-other, or no-hand frames and lets users jump to those moments. A 16-participant study shows no speed gain but a significant increase in perceived ease; users find individual and joint labels helpful, implicit labels and detection accuracy cause false positives and occasional information overload.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 0:17Individual contributions
- 1:00Common approach
- 1:36Question
- 2:20Purpose
- 3:47Interface
- 5:55Demo
- 6:25Evaluation
- 6:52Experiment
- 8:30Results
- 9:05User Feedback
- 9:58Visualization
- 10:42Conclusion
- 11:05Questions
related talks
27mAnalysis of sports scenes using computer vision models
10mProactive Sensing for Improving Hand Pose Estimation
06mMultimodal Joint Head Orientation Estimation in Interacting Groups via Proxemics and Interaction ...
10mGO-Finder: A Registration-Free Wearable System for Assisting Users in Finding Lost Objects via ...
19mEveryday Eye Contact Detection Using Unsupervised Gaze Target Discovery
17mEyeGrip: Detecting Targets in a Series of Uni-directional Moving Objects Using Optokinetic ...
10mHandyTrak: Recognizing the Holding Hand on a Commodity Smartphone from Body Silhouette Images
20mToward Realistic Hands Gesture Interface: Keeping it Simple for Developers and Machines