Multi-modal tooth decay recognition based on Contrastive Learning and Multi-label

In the field of image processing, multi-label learning poses a significant challenge, especially in cavity recognition. Additionally, combining multi-label learning with contrastive learning represents a substantial advancement in image processing and serves as a motivation to support cavity recognition in humans, aiding dentists in making more effective diagnoses. In this paper, we propose a deep learning model to learn both multi-label features and image features. Since multi-label features and image features are inherently discrete, using contrastive learning and a multi-label model is a way to connect them. Extensive experiments were conducted on the ImageNet-mini dataset and our self-collected P-Deltal dataset, achieving 85.65% and 89.28% mAP, respectively, for each dataset.