ml-car-count

Drive-Thru Car Counting with YOLOv8

This project uses the YOLOv8 object detection model to count cars in a drive-thru video by tracking their dwell time within a defined Region of Interest (ROI). It processes video input, detects cars, calculates Intersection over Union (IOU) with the ROI, and outputs an annotated video with car counts.

Features

• Object Detection: Uses YOLOv8 (medium model, yolov8m.pt) to detect cars.
• ROI Tracking: Counts cars that stay within a specified ROI for at least 5 seconds.
• Visualization: Draws bounding boxes (green if IOU > 0.3, red otherwise) and displays car IDs, IOU values, and total count.
• Output: Saves an annotated video with processing metrics.

Prerequisites

• Python 3.7+: Ensure Python is installed.
• Google Colab: Recommended for free GPU access (T4), though it can run locally with a compatible GPU.
• Dependencies:
  • opencv-python (cv2)
  • numpy
  • ultralytics (for YOLOv8)
  • torch (automatically installed with ultralytics)