AI Road Condition Assessment Platform

I need a complete, production-ready system that ingests high-resolution imagery from both drone flights and fixed roadside cameras, uses machine-learning models to identify and quantify each defect, and then turns those results into concise TMH9-compliant reports in both PDF and Excel. The solution must cover every step of the workflow: image acquisition, secure cloud storage, automated preprocessing, model training/inference, geospatial visualisation, interactive dashboards, and one-click report generation.

Key deliverables
• Image-capture pipeline with support for my existing drones and static cameras, plus metadata syncing (GPS, time-stamp, orientation).
• ML models (you may use TensorFlow, PyTorch, Detectron2, etc.) that detect cracks, potholes, rutting and other surface distresses, outputting measurements suitable for the TMH9 rating system.
• Web dashboard (React, Angular or similar) showing a map overlay, defect heat-maps, filtering, and side-by-side image comparisons.
• Reporting engine that exports the analytics to neatly formatted PDF and Excel, ready for client submission.
• API endpoints and documentation so future mobile or GIS tools can call the same functions.
• Deployment scripts (Docker/Kubernetes) and CI/CD so the platform can be installed on our AWS account or an on-prem server.

Acceptance criteria
1. Sample data set processed end-to-end with ≥90 % precision/recall on defect detection.
2. TMH9 grading automatically calculated and visible in the dashboard.
3. A generated PDF and Excel report that matches the sample template I will provide.
4. Source code handed over in a private Git repository with a README that lets a new engineer spin up the entire stack in under one hour.

Please outline your technical approach, estimated timeline, major milestones and any prior work with geospatial imagery or road-surface analytics. I’m happy to answer clarifying questions and can share additional data samples once we shortlist candidates.