ABSTRACT
Real-time traffic object detection is a key topic in computer vision, especially for improving traffic safety and management. This research describes a novel strategy for detecting traffic actors in real-time using YOLOv7, a cutting-edge deep learning system. Traditional computer vision algorithms, such as Single Shot Detector, R-CNN, and older versions of You Only Look Once (YOLO), frequently exhibit slow response times and poor accuracy in high-traffic areas.YOLOv7, an advanced object detection method based on convolutional neural networks (CNNs), is used in the proposed approach to address these difficulties straight on. YOLOv7 not only achieves real-time object detection, but also greatly increases accuracy by removing superfluous candidate boxes and employing a non-maximum suppression module to choose the best bounding boxes from overlapping ones. Furthermore, the spatial pyramid pooling block improves accuracy by enhancing the network’s receptive field without introducing additional parameters. In this study, we demonstrate the performance of our model under various driving scenarios, including clear and cloudy skies, varying lighting, occlusions, and noisy input data. This model detects traffic participants such as automobiles, pedestrians, cyclists, and traffic signs, which contributes to improved traffic safety and management.
