Advanced PTZ Camera Control via Head-Tracking for Hybrid Communications

This paper presents a full-stack solution for controlling a networked PTZ (pan-tilt-zoom) camera via head-tracking and eye-gaze estimation. We detail the evolution of our Python/Flask codebase from a simple Haar-cascade baseline through zoom-and-smooth heuristics to a complete dlib-based solve PnP pipeline with calibration, gaze offsets, dynamic constraints and smoothing. Key algorithms, formulas (e.g. Perspective-n-Point, smoothing, zoom mapping), and calibration strategies are all specified. Experimental results demonstrate accurate, jitter-free tracking in real-time, and we provide extensive discussion of code revisions, modules added, parameters tuned and branches removed.