An Attention Mechanism-integrated Deep Learning-based AI Visual Detection Method for Small Sample Scenarios

Abstract

This study addresses the challenges of weak generalization and category confusion in small-sample AI visual detection caused by
scarce labeled data. By integrating attention mechanisms with deep learning, we propose two core principles: focusing on key features and
enhancing feature distinguishability. The framework introduces an attention-based feature extraction layer and a feature-guided interaction
mechanism. Experimental results demonstrate that this fusion approach effectively reduces feature redundancy, clarifies category boundaries, and improves detection accuracy and model generalization in small-sample scenarios, providing robust support for related technological applications.