The application of computer vision techniques has brought more advancement to traffic safety analysis by allowing researchers to study traffic conflicts from vehicle data. Traffic management systems capture video data and leverage advances in video processing to detect and monitor traffic incidents.
In this work, we design, implement, and analyze a computing model for a smart traffic monitoring system. Moreover, we can use the concept of smart traffic in:
YOLO-NAS is a foundation model for object detection. It improves small object detection, localization accuracy, and performance-per-compute ratio. The "NAS" component is used to automate the design process of neural network architectures. Instead of relying on manual design, we use NAS algorithms to discover the most suitable architecture for a given task. The aim of "NAS" is to find a model that achieves the best accuracy, computational cost, and an optimized model.
The following prototype processes a series of images captured by cameras or videos. It makes predictions on each image, identifying and localizing the Region of Interest (ROI). Detected objects belonging to classes are filtered based on a predefined confidence threshold (e.g., confidence $\ge 0.6$) to ensure reliability.
By utilizing the capabilities of object detection models and advanced analytics, there is a potential to revolutionize urban mobility and enhance road safety in the years to come.
.webp)
