An Intelligent Transportation System (ITS) built using Predictive Analytics and Computer Vision, developed as an academic project at
Lovely Professional University.
This system performs real-time vehicle detection, speed estimation, traffic violation detection, and license plate recognition, all integrated into an interactive Streamlit dashboard with automated reporting.
🔗 View Project Video on LinkedIn
✅ Multi-Class Vehicle Detection
Detects and classifies 7 vehicle classes:
- Car
- Truck
- Bus
- Auto
- Two-Wheeler
- License Plate
- Blurred Plate
✅ Real-Time Speed Estimation
Speed calculated using centroid displacement between frames:
✅ Traffic Violation (Overspeeding) Detection
Automatically flags vehicles exceeding a configurable speed threshold (e.g., 80 km/h).
✅ Automatic License Plate Recognition (ALPR)
OCR-based number plate extraction for violating vehicles.
✅ Strategic Traffic Analytics Reports
Auto-generated PDF reports including:
- Vehicle flow analysis
- Speed & violation trends
- Model confidence & robustness indicators
| Component | Technology |
|---|---|
| Model | YOLOv8 (Nano, Small, Medium tested) |
| Framework | Ultralytics YOLOv8 |
| Language | Python |
| Dashboard | Streamlit |
| OCR | EasyOCR |
| Database | MySQL, MS Excel |
| Visualization | OpenCV, Matplotlib |
Vehicle Detection & License Plate Dataset (v1)
- 📸 ~960 annotated traffic images
- 🏷️ YOLO-format labels
- 🔀 Split:
- Train: 772
- Validation: 127
- Test: 61
Sources:
- Kaggle
- Roboflow Universe
- Image resizing to 640 × 640
- Pixel normalization (0–1)
- Data augmentation:
- Mosaic augmentation
- HSV scaling
- Horizontal flipping
After comparing multiple YOLOv8 variants, YOLOv8n (Nano) was selected.
| Model | Parameters | mAP@50 | Inference Speed | Remarks |
|---|---|---|---|---|
| YOLOv8n | 3.2M | 0.829 | 6.5 ms | ✅ Best for real-time |
| YOLOv8s | 11.2M | 0.841 | 12.8 ms | Good accuracy |
| YOLOv8m | 25.9M | 0.855 | 22.4 ms | Too slow for CPU |
📌 YOLOv8n offers the best balance between accuracy and real-time speed.
- Upload Image / Video
- Live camera feed
- Speed threshold controls
- Real-time annotations
- Downloadable PDF report
📷 Dashboard Preview
(Add image: assets/dashboard.png)
The system generates Strategic Traffic & Safety Reports including:
- 📈 Vehicle flow distribution
- 🚦 Speed & violation analysis
- 🔍 OCR-based license plate records
- 📊 Confidence & robustness metrics
- ✅ Actionable safety recommendations
- 🌙 Night-time & thermal vision detection
- ⚡ Edge deployment (Jetson Nano / Raspberry Pi)
- 🧠 Accident prediction using LSTM
- 🪖 Helmet detection & triple riding detection
- 🚓 Advanced traffic rule enforcement
Sushant Kumar
B.Tech CSE, Lovely Professional University
Registration No: 12311087
Dr. Tanima Thakur
Assistant Professor, CSE/IT
Lovely Professional University