Vehicle Identification from Traffic Video Surveillance Using YOLOv4

Traffic Video Surveillance has emerged as an important tool in our lives. The ability to detect & follow vehicles, both individual & public transport ones, is extremely useful in numerous applications such as the safety of the drivers, preventing & monitoring road accidents etc. Using continuous video stream from CCTV cameras from all over the world, this project deals with the concept of Vehicle Detection with the support of Computer Vision algorithm in real-time frame. The proposed framework capitalizes on YOLOv4 to achieve faster object detection in real time and using the proposed dataset it was tested on various conditions such as rain, low visibility, daylight, snow, and night. Pre-Processing steps of the dataset is done in the first phase. After the first phase, Various types of Vehicle images are captured. In the final phase, depending on the properties computed or calculated in the earlier phases, the Precision, Mean Average Precision (map) and Average Intersection Over Union (IoU) of the model with which it can recognize the various types of Vehicles were found. This model can further help in the development of a framework for Vehicular accident detection in real time.