FIELD NOTE • OSHODI MARKET
"After rain, this drain clogs in 20 minutes. PET bottles from cold-drink sellers + nylon from market. Water enters shops." — Iya Tolu, trader. Empathy first.
● VISION RESULTS • OPENCV 594.3% CONF
1 Preprocess: CLAHE + FastNlMeans Denoise 2 Semantic Segmentation: Waste vs Water vs Concrete 3 Blockage Ratio: Contour + Morphological 4 Risk Classification: Spatiotemporal model
78%
HIGH RISK
arm64 Graviton2 - COOL optimized • Lagos
WASTE COMPOSITION
PET Bottles45%
Nylon / Sachet30%
Organics / Leaves15%
Silt / Sand10%
● AGENTIC REASONING • BEDROCKCLAUDE 3.5 SONNET
01 Perceive: Vision pipeline output blockage=78% > threshold 70% = TRUE
02 Context: Geo-fence Oshodi Market (Lat 6.55 Lng 3.34) • Market density HIGH - School 200m - Rain forecast 80%
03 Reason: IF blockage>70 AND market=TRUE AND PET>40% => Recyclable waste surge, Incentivize pickers. IF water_level>05 + Stagnation risk malaria
DECISION PLAN (Bedrock Agent):
• Action 1: Alert LAWMA Dispatch via SNS - Priority CRITICAL
• Action 2: WhatsApp Community Leader (Iya Tolu + 2) - Clear PET in 45min
• Action 3: Update S3 Digital Twin - GeoJSON flood risk layer
• Action 4: QuickSight push - City dashboard orange alert
1 PERCEIVE Edge + Camera 42ms
2 REASON Bedrock Claude 3.5 1.2s
3 ACT Lambda + S3 280ms
⚠ LAWMA DISPATCH TRIGGERED — CRITICAL Team Oshodi-2 notified — ETA 22 min — Truck 07 — Graviton4 Lambda
Theme: Build vision systems that see, reason, act • COOL Award: Low-cost ARM64