Detecting 3D Hand Pointing Direction from RGB-D Data in Wide-Ranging HRI Scenarios
summary
A mobile-robot pipeline that uses Kinect RGB-D, trains a YOLOv4 hand+human detector on a bespoke COCO-derived dataset, filters pointing hands with a CNN that achieves 94.0% classifier accuracy, segments the hand with an improved color-aware density clustering, and estimates direction from the principal eigenvector of the 3D hand points achieving 83.7% within 20° on a new distance-varied dataset while providing sensor-noise bounds via Weyl and Davis–Kahan analyses.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:42Proposed 3D Pointing Direction Estimation: 3 Steps
- 2:23Hand and Human Detection
- 3:523D segmentation
- 6:10Noise sensitivity
- 6:47Experimental Results: Pointing Gesture Classification
- 7:09Experimental Results: 3D Pointing Direction Data Set
- 8:11Experimental Results: 3D Pointing Direction IPO Data Set
- 8:52Conclusion
related talks
14mBack-Hand-Pose: 3D Hand Pose Estimation for a Wrist-worn Camera via Dorsum Deformation Network
27mAnalysis of sports scenes using computer vision models
07mFingerTrak: Continuous 3D Hand Pose Tracking by Deep Learning Hand Silhouettes Captured by ...
10mProactive Sensing for Improving Hand Pose Estimation
20mToward Realistic Hands Gesture Interface: Keeping it Simple for Developers and Machines
14mDeepFisheye: Near-Surface Multi-Finger Tracking Technology Using Fisheye Camera
05mTouchPose: Hand Pose Prediction, Depth Estimation, and Touch Classification from Capacitive Images
28mDeep Learning-Based Human Activity Recognition for Physical Skills...