A review of small target detection based on YOLO model
Due to the excellent performance of the YOLO model,it has been widely used in many fields such as in-dustry and agriculture.In the process of developing the YOLO series of basic models from YOLOv1 to YOLOv8,the ability of YOLO multi-scale fusion and the optimization of the network structure have been continuously enhanced to improve the model performance.Meanwhile,the other branch models of YOLO,based on the basic models of the YOLO series,have their functions in terms of loss function,backbone network,and multi-task of the basic model greatly improved.This paper mainly discusses the development process of the YOLO model,and studies and discuss-es the accuracy of the YOLO model in detecting small targets in the image and the accuracy of detecting defects in the thangka image to study whether the YOLO model can detect Tibetan cultural relics.For example,whether the YO-LO model can correctly identify and classify Tibetan silverware and copperware,because these two kinds of objects have more styles and less different patterns,requiring a high performance of the model.In this paper,the focus and strategy of the model improvement are discussed and studied,and the YOLO model can find improvement points in the direction of unsupervised or weak supervision in the learning mode.
NETL
NSTL
万方数据
