Vehicle Detection, Tracking, and Traffic Violation Module

An end-to-end, high-performance Computer Vision system designed for real-time vehicle detection, multi-object tracking, traffic light state classification, and automated red-light violation detection in complex traffic video feeds.

Supports multiple tracking engines:

• BoT-SORT: Camera motion compensation + high-precision tracking.
• ByteTrack: Lightweight, ultra-fast low-confidence association.
• DeepSORT: Appearance-based ReID feature embedding and Kalman filter tracking.

---

🌟 Key Features

• 🏎️ Multi-Class Vehicle Detection & Tracking: Detects and tracks 5 distinct vehicle categories (Bicycle, Car, Motorcycle, Bus, Truck) simultaneously using COCO-trained YOLO models.
• 🔀 Multi-Tracker Architecture: Switch seamlessly between BoT-SORT, ByteTrack, and DeepSORT by updating a single configuration setting.
• 🔴 Automated Traffic Light Detection: Real-time traffic signal classification using HSV color space sampling with sliding-kernel peak response and temporal majority voting.
• 🚨 Red-Light Violation Detection: Spatial polygon tracking that continuously monitors vehicle trajectories from approach lanes into intersection regions during red light signals.
• ⚡ Optimized Pipeline: Asynchronous multi-threaded FrameReader decouples CPU video decoding from GPU inference, with FP16 tensor acceleration.
• 🛠️ Interactive Zone Drawing GUI: Built-in OpenCV GUI (ZoneDrawer) to interactively map lane boundaries, intersection polygons, and traffic light coordinates directly on video frames.
• 📊 Headless & Cloud-Ready: Configurable for headless Linux GPU instances (e.g. AWS, ThunderCompute) or local live visual preview HUD.

---

📁 Repository Structure & Modules

VehicleDetectionAndTrackingModule/
├── trackers/                     # Modular tracking engines
│   ├── __init__.py               # Tracker factory `create_tracker(config)` & registry
│   ├── base_tracker.py           # Abstract BaseTracker (interface & common logic)
│   ├── botsort/
│   │   ├── __init__.py
│   │   └── botsort_tracker.py    # Ultralytics YOLO + BoT-SORT
│   ├── bytetrack/
│   │   ├── __init__.py
│   │   └── bytetrack_tracker.py  # Ultralytics YOLO + ByteTrack
│   └── deepsort/
│       ├── __init__.py
│       └── deepsort_tracker.py   # YOLO detector + DeepSORT (deep-sort-realtime)
├── config.py                     # Central configuration & tracker selector
├── vehicle_model.py              # Backward-compatible adapter delegating to trackers/
├── main.py                       # Pipeline orchestrator and FrameReader loop
├── traffic_light_detector.py     # HSV-based traffic light state monitor
├── zone_checker.py               # Point-in-polygon trajectory & violation logic
├── zone_drawer.py                # Interactive OpenCV zone mapping utility
├── visualizer.py                 # Frame HUD renderer, annotations, & statistics overlay
├── setup_cloud.sh                # Environment installer for cloud GPU instances
├── requirements.txt              # Unified Python dependency requirements
├── zones.json                    # Polygon coordinates for lanes, intersections, & lights
├── test_1.mp4                    # Sample input traffic video 1
└── test_2.mp4                    # Sample input traffic video 2

---

⚙️ Architecture & Pipeline Flow

flowchart TD
    A[Input Video File] --> B[FrameReader Background Thread]
    B -->|Decoded BGR Frame| C[Active Tracker<br/>BoT-SORT / ByteTrack / DeepSORT]
    B -->|Decoded BGR Frame| D[TrafficLightDetector HSV Monitor]
    
    C -->|Tracked Bounding Boxes & IDs| E[ZoneChecker Polygon Engine]
    D -->|Light State: RED / GREEN| E
    
    E -->|Violation Events| F[Visualizer HUD Renderer]
    C -->|Track Labels & Boxes| F
    D -->|Light Status| F
    
    F -->|Annotated Frame| G[VideoWriter Output / Live Preview]

Loading

---

🛠️ Installation & Setup

Prerequisites

• Linux OS (Ubuntu 20.04+ recommended)
• Python 3.9+
• NVIDIA GPU with CUDA support (or CPU mode for local testing)

Quick Setup

1. Automated Setup (Cloud GPU / Linux):

   bash setup_cloud.sh

2. Manual Setup:

   # System packages
   sudo apt-get update && sudo apt-get install -y ffmpeg libgl1
   
   # Python dependencies
   pip install -r requirements.txt
   
   # Pre-download YOLO weights
   python -c "from ultralytics import YOLO; YOLO('yolo26l.pt'); YOLO('yolov8l.pt')"

---

🚀 How to Switch Models & Trackers

In config.py, set TRACKER_TYPE to your desired tracker:

# Select active tracker: 'botsort', 'bytetrack', or 'deepsort'
TRACKER_TYPE = 'botsort'

Tracker Configurations

• BoT-SORT:

  BOTSORT_CONFIG = {
      'tracker_yaml': 'botsort.yaml',
      'track_thresh': 0.35,
      'match_thresh': 0.6,
      'track_buffer': 60,
  }

• ByteTrack:

  BYTETRACK_CONFIG = {
      'tracker_yaml': 'bytetrack.yaml',
      'track_thresh': 0.35,
      'match_thresh': 0.6,
      'track_buffer': 60,
  }

• DeepSORT:

  DEEPSORT_CONFIG = {
      'max_age': 60,
      'n_init': 3,
      'max_cosine_distance': 0.2,
      'nn_budget': 100,
      'embedder': 'mobilenet',  # 'mobilenet' or None (for IoU-only)
      'half': True,
      'embedder_gpu': True,
  }

---

🏃 Running the Application

1. Run Pipeline

python main.py

2. Interactive Zone Drawer GUI

To configure lanes, intersection zones, and traffic lights:

1. In config.py, set:

   ENABLE_ZONE_DRAWER = True

2. Run:

   python main.py

3. Controls:
   • 1: Start drawing Lane polygon (Click points on frame)
   • 2: Start drawing Intersection polygon (Click points on frame)
   • 3: Click to set Green Light coordinate
   • 4: Click to set Red Light coordinate
   • c: Complete current polygon
   • s: Save all zones to zones.json
   • r: Reset active points
   • d: Delete last added zone/light
   • ESC: Exit drawer
4. Set ENABLE_ZONE_DRAWER = False to resume video tracking.