Unlabeled Data Augmentation with Diffusion Model for Semi-Supervised Object Detection

In the field of medical object detection, acquiring abundant high-quality data remains a critical challenge. To address this issue, we propose DABDM (Data Augmentation Based on Diffusion Model), a semi-supervised object detection framework that exploits diffusion models to enhance medical image datasets. Our approach generates a large volume of synthetic images, which are utilized for the unsupervised training phase in semi-supervised learning, thereby mitigating the reliance on extensive labeled data. Experimental results demonstrate that DABDM significantly improves the performance of medical object detection models, showcasing its potential to advance the field by providing a robust solution to the data scarcity problem. By incorporating the generated unlabeled training data into the semi-supervised framework, we observed a notable improvement in model accuracy. Specifically, our experiments showed an increase of up to 6.92\% after adding the generated images.