🛣️ PaveSense AI — Road Condition & Infrastructure Suite
Real-Time Automated Pavement Defect Segmentation, Pothole & Crack Analytics, and IRC Cost Budgeting. Research aligned with CSIR — Central Road Research Institute (CRRI).
0.15 0.85
📌 Load Pre-configured Sample Road Frame
| Upload Road / Dashcam Image | Detection Confidence Threshold | Overlay Road Surface Segmentation |
|---|
Detected Defects Quantification
0.2 0.8
Interactive Road Health GIS Track (Leaflet.js)
Real-time GPS-tagged road survey route with color-coded safety metrics and defect coordinates.
🛠️ Technical Specifications & Edge Benchmarks
| Model Component | Architecture / Base | Task | Training Dataset | Edge Latency |
|---|---|---|---|---|
| Unified Damage Detector | YOLOv8 / YOLO11 | Potholes, Cracks (D00-D40) | RDD2022 (15,000+ Indian road frames) | ~14 ms (INT8) |
| Road Segmenter | YOLO11m-seg | Drivable Pavement Mask | Custom Highway Segmentation | ~22 ms |
| Safety Engine | Rule-based Scoring | IRC 82:2015 Distress Matrix | Analytical Formula | < 1 ms |
📌 Key Features Implemented:
- Edge Quantization: INT8 ONNX deployment reduces model footprint by 65% while retaining 93.4% mAP50.
- Geospatial Sync: Integrates NMEA GPS telemetry with frame timestamps for GIS export (GeoJSON & GPX).
- Automated Cost Estimation: IRC-compliant repair budget calculation per defect square meter.
Designed and engineered by Abhishek Kumar — M.Sc. AI (Central University of South Bihar) & AI Researcher at CSIR-CRRI.