poorvika4566/YOLOv8-Traffic-Sign-Light-OCR

YOLOv8 Traffic Sign & Light Detection with OCR

A YOLOv8-based computer vision model for detecting traffic signs and traffic lights from images, video, or live webcam input.

The project also integrates Tesseract OCR to read speed-limit values from detected speed-limit signs.

Note: This is a computer-vision demonstration and is not a certified road-safety system. It should not be used to control a vehicle or make safety-critical decisions.

Model

• Architecture: YOLOv8
• Task: Object Detection
• Model: YOLOv8n
• Input size: 640 × 640
• Training epochs: 10
• Batch size: 16
• Model weights: best.pt

Detection Pipeline

The system follows this process:

1. Capture frames from a webcam or video.
2. Run YOLOv8 object detection.
3. Detect traffic lights and traffic signs.
4. Identify speed-limit sign detections.
5. Crop the detected speed-limit sign.
6. Apply image preprocessing.
7. Use Tesseract OCR to read the speed value.
8. Validate the OCR result against accepted speed-limit values.
9. Display the detected sign and recognized speed.
10. Save detection results to a CSV file.

OCR

Tesseract OCR is used as a second verification step for speed-limit signs.

The current implementation accepts the following speed values:

5, 10, 15, 20, 25, 30, 60, 80