Research of target detection method YOLO

Abstract

Regression theory is used by YOLO technology to build the one stage detection technique. It merely employs a trunk CNN to predict various targets using the "feature extraction-direct regression" method. In comparison to previous algorithms, it detects things much faster and with far higher accuracy. Researchers have become interested in the YOLO model because it is widely employed in sectors such as autonomous driving, camera display, video surveillance, vehicle identification, face recognition, remote sensing satellite, infrared detection, and others. In this study, the model structural properties of the yolo model and the yolov1-v7 models are primarily analyzed and contrasted. According to the model design perspective, it compares and summarizes the structural enhancement and corresponding performance optimization of each model in the model structure, assesses the benefits and drawbacks of each model, and assesses their performance in light of the actual application impact of each model and the primary application fields, serving as a reference for the study of related topics. The article's summary and future direction are provided at the end.