ParkLane - License Plate Detection

Smart parking system that detects and reads vehicle license plates using computer vision.

About

ParkLane is a system where apartments and malls can let users book parking spots online. A camera at the entrance detects the car's number plate automatically when it arrives.

This repo has the number plate detection part. Two approaches:

• OpenCV - basic contour detection (detect.py)
• YOLOv8 - trained model for better accuracy (detect_yolo.py)

Setup

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Method 1: OpenCV (basic)

python detect.py

Uses edge detection and contour finding to locate plates. Works okay but misses some plates.

Method 2: YOLOv8 (better)

First prepare the dataset and train:

python prepare_dataset.py
python train.py

Then run detection:

python detect_yolo.py

Or test on a single image:

python detect_yolo.py dataset/images/some_image.png

YOLOv8 is way more accurate than the contour method.

Dataset

Using the Car License Plate Detection dataset from Kaggle (433 images).

Tech Stack

• Python
• OpenCV
• EasyOCR
• YOLOv8 (ultralytics)