Res-GeoAI: A Satellite-to-Drone Dual-Modality System for Flood Detection and Victim Localization in Disaster Response

---

Presented at: CVIP 2025 (10th International Conference on Computer Vision and Image Processing 2025)

Rapid and effective disaster response remains a critical challenge, particularly during floods, where timely identification of affected areas and victims can save lives. Existing systems often rely solely on either satellite imagery or ground-based sensors, limiting their ability to quickly and accurately pinpoint flooded zones and locate stranded individuals.

To address this gap, we propose Res-GeoAI, a geospatial AI system that integrates satellite-based flood detection with drone-assisted victim identification. Our system operates in three hierarchical stages:

1. Satellite Flood Mapping — Semantic segmentation of satellite imagery to detect flooded zones
2. Flood Validation — CNN-based classification to validate candidate flood regions
3. UAV Victim Detection & Localization — Thermal imaging and object detection for victim identification with GPS coordinates

By combining these capabilities, Res-GeoAI enhances the speed and precision of humanitarian interventions, providing a scalable solution for disaster management.

---

System Overview

Figure 1. End-to-end system workflow for flood victim detection and localization. Level 1 (top): satellite segmentation and flood coverage analysis across large geographic regions. Level 2 (bottom): UAV victim detection via thermal imaging and GPS localization for coordinated rescue operations.

---

Experimental Results

1. Semantic Segmentation — LoveDA Dataset

Comparison of segmentation models for 7-class land-use classification on satellite imagery:

Model	Background	Building	Road	Water	Barren	Forest	Agriculture	mIoU
UNet	43.46	60.59	52.35	69.22	46.60	24.23	41.88	48.95
Semantic FPN	60.70	50.92	51.36	73.93	52.05	56.24	76.94	57.98
FuseNet	47.19	50.77	61.36	68.21	31.71	47.71	57.45	52.62
HRNet	54.61	55.34	57.42	73.36	46.78	45.87	69.88	59.40
Ours (CNN+SegFormer)	60.68	63.25	62.08	76.47	47.93	47.82	72.75	66.30

Our modified CNN+SegFormer architecture achieves the highest mIoU (66.30%) with superior water class detection (76.47%), critical for flood mapping.

2. Flood Validation — FloodNet Dataset

Binary classification accuracy for flood region validation using Xception-based classifier:

Model	Training Accuracy	Test Accuracy
InceptionNetV3	0.990	0.844
ResNet50	0.974	0.937
Xception (Ours)	0.998	0.946

The Xception-based validator achieves 94.6% test accuracy, effectively reducing false positives in flood detection and triggering UAV deployment only when confidence exceeds 0.7.

3. Victim Detection — HIT-MOB Dataset (Multi-Modality)

VFNet victim detection performance across different imaging modes (RGB, Thermal, Infrared, Night Vision):

Model	RGB	Thermal	Infrared	Night Vision
YOLOv8 (mF1 / MPR)	0.841 / 0.136	0.846 / 0.136	0.850 / 0.133	0.857 / 0.131
VFNet (mF1 / MPR)	0.858 / 0.117	0.863 / 0.114	0.860 / 0.115	0.865 / 0.111

Night vision mode achieves the highest mF1 score (0.865), optimal for low-visibility disaster scenarios with improved victim detection and minimal false alarms.

4. System Performance

End-to-end pipeline execution metrics on a 1000 km² region (Zoom Level 11):

Module	Input Size	Execution Time	Memory Usage	Accuracy
Tile Generation	1024×1024	0.80s	17.5 MB	1.00
Flood Detection	1024×1024	0.05s	4.5 MB	0.66
Flood Validation	1024×1024	0.02s	1.5 MB	0.95
Victim Detection	512×512	0.04s	2.0 MB	0.86
GPS Localization	—	0.01s	0.1 MB	±3.2m
Total	—	0.92s	25.6 MB	Efficient

---

Repository Structure

Res-Geo-AI/
├── main.py                        # Entry point for full pipeline
├── requirements.txt               # Python dependencies
├── README.md                      # This file
├── configs/
│   ├── config.py                  # Dataclass config + YAML loader
│   └── default.yaml               # Default hyperparameters
├── satellite/
│   ├── tessellation.py            # MBR, grid tessellation, tile filtering
│   ├── segformer.py               # Modified SegFormer with multiscale attention
│   ├── flood_detection.py         # Per-tile flood inference and coverage
│   ├── flood_validation.py        # Xception-based flood region validation
│   └── pipeline.py                # Satellite pipeline orchestrator
├── uav/
│   ├── night_vision.py            # RGB → night vision / thermal / IR transforms
│   ├── varifocalnet.py            # VFNet with Res2Net backbone + varifocal loss
│   ├── victim_detector.py         # Inference, NMS, person-class filtering
│   ├── flight_planner.py          # Spiral UAV path planning
│   └── pipeline.py                # UAV pipeline orchestrator
├── fusion/
│   └── pipeline.py                # Satellite–UAV feature fusion and score merging
├── gps/
│   ├── localization.py            # Pixel→GPS projection, clustering, error propagation
│   └── kalman_filter.py           # 6-DOF Kalman filter for pose smoothing
├── utils/
│   ├── logger.py                  # Centralized logger factory
│   ├── io.py                      # Image loading, JSON I/O, overlay saving
│   ├── metrics.py                 # IoU, mIoU, F1, MPR, GPS error calculation
│   └── visualization.py           # Segmentation colorization, cluster plots
├── dataset/
│   └── __init__.py
├── images/
│   └── fig1.png                   # System workflow diagram
└── testing/
    ├── run_all_tests.py           # Test suite runner
    ├── test_segformer.py          # SegFormer architecture tests
    ├── test_xception.py           # Flood validator tests
    ├── test_vfnet.py              # VFNet victim detector tests
    ├── test_gps.py                # GPS localization tests
    ├── test_tessellation_and_planning.py
    └── saved_models/              # Place .pth checkpoints here
        ├── segformer_flood.pth
        ├── xception_flood.pth
        └── vfnet_res2net101.pth

---

Quick Start

Installation

# Clone the repository
git clone https://github.com/Anidipta/Res-Geo-AI.git
cd Res-Geo-AI

# Install dependencies
pip install -r requirements.txt

Usage Examples

# Full end-to-end pipeline (satellite + UAV + GPS)
python main.py --region data/region.geojson --output outputs/ --mode full

# Satellite-only mode (flood detection & validation)
python main.py --region data/region.geojson --output outputs/ --mode satellite

# UAV victim detection mode (from pre-detected flood regions)
python main.py --flood-tiles outputs/flood_tiles.json --output outputs/ --mode uav

# Run complete test suite
cd testing && python run_all_tests.py

---

Citation

If you find this work useful in your research, please cite:

@article{pal2025resgeoai,
  title     = {Res-GeoAI: A Satellite-to-Drone Dual-Modality System
               for Flood Detection and Victim Localization
               in Disaster Response},
  author    = {Pal, Anidipta},
  year      = {2025},
  url       = {https://doi.org/10.13140/RG.2.2.32814.88642}
}

---

Contact & Resources

• Author: Anidipta Pal
• GitHub: Anidipta/Res-Geo-AI
• Dataset: HIT-MOB on Kaggle
• Preprint DOI: 10.13140/RG.2.2.32814.88642