ANANSI v1 — Nigeria Sovereign Terrain Intelligence Model
Developed by: Krystallx Shield · Engr. Ahmadu Yusuf Soje
Classification: Sovereign Development Asset
Territory: Federal Republic of Nigeria
Last Updated: States trained: Rivers, Borno
Training Coverage
- States trained: 2 / 37 (36 states + FCT)
- Total chips: ~4,132
- Completed states: Rivers, Borno
- Remaining states: 35 to go
Architecture
- Base: ViT-Base (Vision Transformer, 86M params)
- Fine-tuned classification head for 6 Nigerian terrain classes
- Input: Sentinel-2 L2A RGB (B02, B03, B04) @ 224×224px @ 10m
Terrain Classes
| ID | Class | Description |
|---|
| 0 | Urban/Built-up | Roads, buildings, settlements |
| 1 | Vegetation/Savanna | Guinea savanna, shrubland |
| 2 | Water | Rivers, lakes, creeks |
| 3 | Bare Soil/Laterite | Exposed soil, mining sites |
| 4 | Agriculture | Farmlands, crop fields |
| 5 | Forest | Dense forest, mangroves |
Intended Use
Sovereign defense intelligence, illegal mining detection, flood mapping,
encampment detection, and geospatial threat analysis for Nigerian
government and defense institutions via the Krystallx Shield C4ISR platform.
