Geospatial Intelligence Platform
MODEL ACCURACY · mIoU
vs baseline 0.75–0.85
0.9566
Village Selection
AI Processing Pipeline
1
Loading GeoTIFF
2
Tiling 512×512 patches
3
SegFormer MIT-B2 inference
4
Generating vector polygons
5
Preparing visualization
Model Specification
SegFormer · MIT-B2 Backbone
Parameters27.4 M
Training patches198,678
Classes11
Input size512 × 512 px
Best epoch3 / 4
Our mIoU0.9566 ✓
Baseline: 0.75–0.85+12% above
🌐 SELECT A VILLAGE · CLICK LOAD DEMO TO BEGIN
ANALYZING TERRAIN...
LAT: — | LNG: —
Layer Controls
Buildings
Roads
Water Bodies
Infrastructure
Map View Mode
📷RAW
🛰SAT + AI
🗺MAP + AI
Detection Statistics
Buildings
Roads
Water
Total Polygons
Class Distribution
Legend
Buildings (Built-up Area)
Roads / Centre Lines
Water Bodies
Infrastructure / Utility
Export Results
System Info
CG villagesEPSG:32644
PB villagesEPSG:32643
Output CRSPreserved ✓
Inference GPURTX 3050
Chunk size4096 × 4096 px