Comparison and Application of Vehicle Target Detection Methods Based on YOLOv5s and YOLOv7

Currently, the methods for pedestrian detection have undergone significant expansion, yet vehicle target detection still retains immense potential for widespread application in the field of intelligent recognition. Nevertheless, vehicle target detection encounters challenges in terms of detection effectiveness, impacted by factors such as lighting conditions, environmental variations, and occlusions. This study centers on vehicle target detection, aiming to tackle the issues of difficult detection, easy missed detection, and algorithm selection within complex backgrounds. By utilizing the YOLOv5s and YOLOv7 algorithms, we have constructed a dataset that encompasses images featuring vehicles against complex backgrounds and in small scales. This dataset is then used to train the models, which are subsequently employed to perform target recognition for vehicles individually. Through comparative experimentation, we have proven the superiority of the YOLOv7 algorithm in detecting vehicles amidst complex backgrounds and those of small scales.