Skip to content

V1.1

Latest
Latest

Choose a tag to compare

View all tags
@RezaGooner RezaGooner released this 03 Aug 22:43
· 7 commits to main since this release
V1.1
f076610
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

🚀 Version 1.1 – Improved Car Name Recognition Model

Multi-Task Vehicle Monitoring System

This is the second stable version of the Multi-Task Vehicle Monitoring System:

A versatile and real-time tool for Vehicle Recognition, Model Recognition, Color Recognition, Persian License Plate OCR, City Extraction and Accident Detection in streaming videos, images or webcam feeds.

Built on a strong deep learning backbone (YOLOv8/YOLOv11, ResNet18, PyTorch) and optimized OpenCV pipelines, the system provides accurate and multi-modal vehicle analysis tailored to Iranian traffic and datasets.

🔑 Key Features

  • Vehicle Recognition (YOLOv11)

Accurate recognition of Iranian vehicles in various scenarios, day and night.

  • Car Brand and Model Classification (ResNet18)

Predicts the class among 27 supported Iranian car models/brands and displays it above each bounding box.

  • Car Color Recognition (ResNet18, 12 colors)

Distinct color classification is presented as colored borders and text on the recognition boxes.

  • Persian License Plate City OCR and Search

Reads Persian numbers and characters, finds the city of registration via internal database.

  • Crash Detection (Collision Detection System)

Real-time crash inference; draws a bold red box, issues an audio alert (beep.mp3).

  • Multiple input modes:

  • Image file

  • Video file

Webcam / Live camera (Desktop only)

  • Real-time FPS display
    The frame processing speed is continuously reported in the output video.

  • Output saving:
    Annotated images/videos are automatically saved with timestamped filenames in the output/ directory.

Easy configuration:
PATHS, THRESHOLDS and IMG_SIZE are modularized for quick setup and expansion.

🚗 Supported vehicle classes

  • 27 Iranian car models/brands:

  • Full list of classes is available in the README and code.

📊 Model Accuracy and Criteria

  • Model: car_name_model.pth (ResNet18, 27 classes)

  • Exact results in README section "📈 Model Evaluation"

📝 Dataset

  • Image/Video Source:

  • Iranian Market Platforms: IranKhodro, Saipa, Divar, Bama

  • Public Dataset via Roboflow

  • City License Plate Data:

Search provided via Plates/city_plateinfo.txt

🔊 Sound Alerts

  • Accident alerts via sound/beep.mp3 (open, replaceable).

🖥️ Installation

Requirements:

  • Python ≥ 3.8
  • PyTorch (CUDA optional but recommended for live FPS)
  • OpenCV, Timm, PaddleOCR, NumPy and other dependencies listed in requirements.txt

Quick Installation:

  • Place all pre-trained weights in weights/ folder (see README)

💡 Usage Guide

1. Image:

  • Set PATHS["source"] = "yourimage.jpg"
  • Output saved in output/

2. Video:

  • Set PATHS["source"] = "yourvideo.mp4"
  • Play or process via GUI; output saved automatically

3. Webcam/Live Streaming:

  • Set PATHS["source"] = 0 (USB camera index)
  • Real-time detection with optional display (cv2.imshow)
  • Exit: Press Q

🏁 Sample Outputs

  • Annotated sample images and videos (see the "Sample Outputs" section in the README ) including:

  • Overlapping vehicle detection

  • License plate and city identification

  • Color-coded vehicle detection

  • Collision inference

🤝 Contribution

Pull requests welcome!

  • Fork the repository
  • Create your feature branch (git checkout -b feature/MyFeature)
  • Commit and push
  • Open a PR

Github.com/RezaGooner

Assets 2
Loading