Human-in-the-Loop Hard Attention for Privacy-Aware Active Vision

This project explores hard attention as an interface for privacy-aware and human-controllable visual recognition. Instead of exposing an entire image to the model, the system processes selected visual glimpses and makes the observation process inspectable.

The project studies how humans may intervene in or audit the visual evidence used by an AI model, connecting active vision with human-AI interaction, privacy-aware AI, and interpretable machine learning.

Keywords: human-in-the-loop AI, privacy-aware vision, active vision, hard attention, visual evidence audit.