Automated 3D Conflict Damage Assessment & Visualization System
Monocular depth estimation pipelines for urban change detection and conflict damage assessment
Overview
Developed an Automated 3D Damage Assessment System using monocular depth estimation model to detect changes in 3D for urban and conflict zone damage assessment. The primary study areas focused on the Ukraine-Russia border regions, including heavily impacted cities such as Marinka and Mariupol.
Project Context
At Data Design Engineering, I engineered end-to-end pipelines for damage assessment using monocular imagery from UAVs and satellites.
Technical Architecture
Input RGB Image ──► Depth Estimation Model ──► Change Detection Model ──► Web Dashboard
Key Features
Monocular Depth Estimation
- Implemented state-of-the-art depth models (Depth-Anything)
- Generated depth maps for change detection
Change Detection Model
- Implemented state-of-the-art change detection models (Bandon)
- Generated change maps for change detection
Web-Based Dashboard
- Deployed via Nginx for remote access
- Interactive visualization of damage areas in 3D
- Labeling interface for damage assessment
Applications
| Application | Description |
|---|---|
| Urban Compliance | Detection of unauthorized construction |
| Conflict Assessment | Structural damage quantification |
| Disaster Response | Post-event infrastructure evaluation |
| Asset Management | Inventory and monitoring of assets |
Technologies Used
| Category | Tools |
|---|---|
| Depth Models | Depth-Anything, Bandon |
| 3D Processing | Open3D, Three.js |
| Deployment | Nginx, Docker, FastAPI |
| Languages | Python, JavaScript |