eSports Gaze Behavior Detection
few-shot gaze tracking for personalized eSports assistance
Undergraduate project (Feb 2022 - Apr 2023).
Overview
A personalized eSports assistant using deep learning for gaze tracking across video frames.
Key Components
- Cross-frame gaze estimation: a model using CNNs and temporal feature alignment
- Few-shot learning: a novel framework allowing rapid adaptation to users with minimal calibration data
- Real-time inference: the final system operates with <50ms latency on standard GPUs
Gaze tracking demo. To be replaced with a figure from the project.