Research on Few-Shot Object Recognition

Object recognition is a crucial task in the realm of computer vision, aimed at identifying and pinpointing a particular object within an image. In practical scenarios, a lack of sufficient training samples frequently hinders the training of deep learning models, presenting a significant challenge. FSOR was created to address this issue. In this paper, the basic concepts, main methods, challenges and latest research progress of few-shot object recognition are reviewed, and its potential in practical application is discussed.