Abstract

We propose and evaluate three algorithmic iterations for detecting and tracking invariant objects (e.g., computer cursors) across video frames. Each edition targets distinct computational budgets and research constraints.

Approach

  • Classical image processing baseline for rapid detection.
  • Feature‑based tracking for robustness under occlusion and motion blur.
  • Hybrid pipeline balancing accuracy and real‑time performance.

Results

  • Consistent frame‑rate on consumer hardware with tunable quality levels.
  • Improved precision/recall versus single‑method baselines.

Publication

Presented at HCII 2024 and URCA 2024. Contact for manuscript.