This implementation for detecting, tracking and counting cars, I’ve labeled Dataset of more than 300 images of cars by top-view using Roboflow. Then by using augmentation features for these images, I’ve created over 3000 images ready for training by YOLOv5. The trained model can detect the cars with a top view, and those in the edges of the video that are partly displayed.
For tracking the cars, I’ve used Deep-Sort Algorithm which uses the Kalman Filter to track the detected cars, then by storing the IDs of the tracked cars in a list, we can count them.

Leave a Reply

Your email address will not be published. Required fields are marked *