Makerere University • Smart City AI

Proactive Urban Flood Management Through Spatio-Temporal Graph Reasoning

Engineered at Makerere University. Drain is a multi-modal, predictive AI infrastructure that abandons reactive municipal maintenance in favor of real-time visual detection and graph-based flood propagation forecasting.

Edge AI
ST-GNN
Uganda-Built
Privacy-Preserving
320 Images Curated dataset of Kampala drainage channels
3.2M Parameters Ultra-lightweight YOLOv8-Nano model
4 Foundational Pillars Vision, Semantic, Generative, Graph
About the Platform

What is Drain?

Drain is an intelligent urban infrastructure platform built for municipal authorities and environmental engineers. It integrates edge computing and cloud-based deep learning into a centralized Web-GIS dashboard. By fusing lightweight object detection (YOLOv8-Nano) operating at the edge via drones and cameras, semantic understanding (CLIP), and Spatio-Temporal Graph Neural Networks (ST-GNN), the system does not just detect a blocked drain—it predicts how that localized blockage will cascade and flood the interconnected city grid.

Municipal Authorities

Automated alerts and dashboard visualizations for maintenance optimization

Emergency Response Agencies

Data-driven foresight into flood propagation and risk mitigation

Smart City Contractors

Scalable API for integrating environmental telemetry into civic platforms

Lightweight Edge Vision

YOLOv8-Nano for real-time detection on drones and cameras

Semantic Generalization

CLIP alignment for deep contextual understanding

Generative Augmentation

GANs and Diffusion for synthetic training data

Spatio-Temporal Graph

ST-GNN for flood propagation forecasting

The Problem We Solve

Recurring Urban Flooding

Drainage channels choked with solid waste, silt, and vegetation cause catastrophic surface flooding in rapidly expanding cities.

Reactive Maintenance

Current maintenance relies on costly, inefficient manual inspections that only identify blockages after flooding has occurred.

Black-Box Sensors

IoT sensors measure water levels without identifying the cause of obstruction, providing no actionable intelligence.

Why Uganda, Why Now

Our Strategic Market Opportunity

In Kampala and across Sub-Saharan Africa, improper urbanization and poor solid waste management have turned seasonal rains into severe economic and humanitarian threats. Addressing this cannot rely on importing expensive, pre-packaged sensor networks designed for Western topographies.

By engineering Drain locally, we operationalize a "Buy Uganda, Build Uganda" philosophy. We provide a frugal, highly scalable infrastructure that empowers local authorities to manage climate adaptation and urban resilience using homegrown computational intelligence, ensuring data sovereignty and context-aware public administration.

Locally Engineered Frugal & Scalable Data Sovereign Climate Resilient

Impact at a Glance

100% Reactive to proactive shift
3.2M Parameters for edge inference
320 Localized images captured
4 Foundational AI pillars

Founders & Strategic Leads

Nabbumba Margaret

Project Visionary, Team Lead & ML Expert

Specializing in YOLO architectures, generative data augmentation, and system pipeline design. Leads the engineering of the Drain detection models.

Ssozi Gloria Edith

Deployment Lead & AI Engineer

Specializing in Spatio-Temporal Graph Neural Networks, MLOps, edge deployment, and Web-GIS dashboard integration.

Dr. Ggaliwango Marvin

Strategic Advisor & AI Mentor

Providing oversight on model optimization, research-to-product commercialization, and responsible AI governance.