NeckFace: Continuously Tracking Full Facial Expressions on Neck-mounted Wearables
summary
A neck-mounted IR camera system reconstructs continuous full-face expressions by feeding chin and mouth images into NeckNet based on ResNet34 and regressing 52 blendshape parameters plus three head angles, trained against Apple TrueDepth/ARKit as ground truth. Hardware comes as a single-camera necklace or two-camera neckband; a 13-participant study tests sitting, walking, and remounting. Reported error is quantified by MAE and authors flag MAE < 40 = visually similar to ground truth.
AI-written summary · about
related talks
07mFaraPy: An Augmented Reality Feedback System for Facial Paralysis using Action Unit Intensity ...
17mHow Cephable Applies On-Device AI to Accessibility
19mEveryday Eye Contact Detection Using Unsupervised Gaze Target Discovery
10mSignGlass: First-Person View Comprehensive and Generalizable ASL Translation Using Wearable Glass
10mTeethTap: Recognizing Discrete Teeth Gestures using Motion and Acoustic Sensing on an Earpiece
14mTransform intent into action through your eyes
21mA Face Recognition Application for People with Visual Impairments: Understanding Use Beyond the Lab
18mOrbits: Gaze Interaction for Smart Watches using Smooth Pursuit Eye Movements