392253 Project: Head Gesture Detection Using Optical Flow (Pj) (WiSe 2026/2027)

Kein Kommentar vorhanden

Teilnahmevoraussetzungen, notwendige Vorkenntnisse

Background and Motivation Nodding and head shaking are fundamental non-verbal communication signals, yet automatically detecting such gestures from unconstrained video remains an open problem. Building on the optical flow-based nodding detection approach of Ghourabi et al. (2020), this project develops an interpretable and computationally lightweight detection pipeline as an alternative to deep learning approaches. A gesture segmentation system will be built using dense optical flow extracted from face regions in naturalistic video to detect nodding and shaking gestures. The resulting system will be evaluated against manual annotations, providing practical insight into classical motion-analysis techniques for head gesture recognition. Skills and Requirements Background in computer vision or image processing. Familiarity with Python and OpenCV. Strong analytical and problem-solving skills. Reference Ghourabi, A., Ghazouani, H., & Barhoumi, W. (2020, September). Driver drowsiness dete