OddEyeCam: A Sensing Technique for Body-Centric Peephole Interaction using WFoV RGB and NFoV ...
summary
A prototype mobile tracking system fuses a wide-view fisheye RGB image with a narrow-field depth camera to estimate device position relative to the trunk and works while standing and walking, validated against an OptiTrack rig. The implementation uses 180° fisheye + RGB-ToF depth camera fusion and body keypoint alignment to a trunk-centered frame, yields 4.17 and 4.47 cm errors (standing, walking) and supports peephole selection that follows Fitts’ law with 7.7% / 18.8% error rates, showing practical, continuous body-centric tracking.
AI-written summary · about
chapters (17)
- 0:00Start
- 0:30Small Display of Mobile Device
- 0:44How to Overcome it?
- 1:14Technical Approaches
- 3:11Design Goals
- 4:15Overall Pipeline
- 6:50Tracking Accuracy Test
- 7:03Covered Wide Space
- 7:21Covered Static + Mobile Env.
- 7:36Covered Various Garments
- 7:47Average Tracking Error: 4.3 cm
- 8:47Design Goal Evaluation on Six Interaction Scenarios
- 10:38Walking detection
- 11:02Results
- 12:04Quantitative Characterization of OddEyeCam
- 12:28OddEyeCam followed Fitts' Law
- 13:13Summary
related talks
20mBuilding a Personalized, Auto-Calibrating Eye Tracker from User Interactions
15mMeCap: Whole-Body Digitization for Low-Cost VR/AR Headsets
10mEllipses Ring Marker for High-speed Finger Tracking
19mEyeScout: Active Eye Tracking for Position and Movement Independent Gaze Interaction with Large ...
07mFingerTrak: Continuous 3D Hand Pose Tracking by Deep Learning Hand Silhouettes Captured by ...
18mCamTrackPoint: Camera-Based Pointing Stick Using Transmitted Light through Finger
17mMadEye: Boosting Live Video Analytics Accuracy with Adaptive Camera Configurations
06mInSight: Monitoring the state of the driver in low-light using smartphones