Vehicle Detection and Counting in Secure Valley Parking Using CNN and YOLO

When trying to manage vehicles in a car park or on a roadside, there can often be issues with keeping the number of vehicles consistent. This project uses computer vision to try to detect and count these vehicles. A YOLOv8 model was trained to detect vehicles. A unique dataset containing 5 classes (Ambulance, Bus, Car, Motorcycle and Truck) was used to train the model. The model was trained using Google Collab with a Tesla T4 GPU. The model was tested with images and videos. Object tracking was added for video detection. An easy-to-use interface was designed. The results show that the model is able to recognise vehicles of various types. However, there are some false positives and issues with the classification. This project demonstrates how deep learning can be used for vehicle monitoring. The groundwork has been laid to increase automated parking and traffic management.