자기지도학습과 비쥬얼 트랜스포머 기반의 얼굴 인식에 관한 연구

Abstract

In this paper, we propose an enhanced face recognition algorithm leveraging self-supervised learning and visual transformer. Visual transformers have demonstrated superior performance in various computer vision tasks. However, they typically require extensive datasets for stable training. To address this limitation, we integrate self-supervised learning techniques with the visual transformer architecture and subsequently fine-tune it to enhance face recognition capabilities. Experimental results show that our proposed algorithm outperforms conventional visual transformer-based approaches in accuracy.