FingerTrak: Continuous 3D Hand Pose Tracking by Deep Learning Hand Silhouettes Captured by ...
summary
A wristband prototype with four 32×24 thermal cameras and a ResNet34-based deep network reconstructs 20 joint 3D hand poses from wrist silhouettes only, using Leap Motion for ground truth. User study with 11 participants reports average angular error of 6.46° same-background, 8.06° different-background, and reasonable performance after re-mounting; three cameras approach four-camera accuracy and RGB images improve fine micro-finger classification versus thermal.
AI-written summary · about
chapters (8)
- 0:00Introduction to FingerTrack
- 1:25System intuition and design
- 2:16System architecture
- 3:01Demos and use cases
- 3:37Evaluation and test scenarios
- 4:23Camera configuration analysis
- 5:30Camera type evaluation
- 5:59Conclusion and applications
related talks
14mBack-Hand-Pose: 3D Hand Pose Estimation for a Wrist-worn Camera via Dorsum Deformation Network
14mDeepFisheye: Near-Surface Multi-Finger Tracking Technology Using Fisheye Camera
10mHandyTrak: Recognizing the Holding Hand on a Commodity Smartphone from Body Silhouette Images
05mTouchPose: Hand Pose Prediction, Depth Estimation, and Touch Classification from Capacitive Images
10mProactive Sensing for Improving Hand Pose Estimation
10mEllipses Ring Marker for High-speed Finger Tracking
11mEfficient Finger Model and Accurate Tracking for Hover-and-Touch, Mid-air and Microgesture Intera...
20mToward Realistic Hands Gesture Interface: Keeping it Simple for Developers and Machines