Project SBAFN is an explainable, street-level flood-proneness visualization and decision-support app for Philippine cities. It visually communicates per-street, evidence-backed flood insights that planners can act on.
Note
This project aims to supersede Project NOAH, the Philippines' state-of-the-art initiative that provides area-based hazard overview, highlighting flood hazard zones at the national level. SBAFN offers the next step by covering the existing gaps of NOAH, by showing the street-level specifics of why floods happen, and give the explainability that was long needed for actual, effective decision making.
- Street-Level Granularity - scores every road segment (green/yellow/red) instead of broad barangay/city polygons.
- Explainability - shows why a specific street scores that way. It includes insights about street-level physical flood proneness indicators, topography data, road data, all mapped to historical flood events as basis.
- Scenario + Prioritization - stress-test each street segment based on previous reported rainfall events (30/50/100 mm/hr rain).
- Actionable Insights - Upon full release of Project SBAFN, outputs can be used for LGU planning, work orders, and DRRM briefings.
➡️ For the specifics, refer to the overview in the Methodology section below or docs/METHODOLOGY.md.
The latest version of Project SBAFN can be accessed through the following link:
https://project-sbafn.vercel.app
To setup this project locally, ensure you have:
| Requirement | Version |
|---|---|
| Python | 3.11+ |
| Git | 2.30+ |
| Flutter | 3.35+ |
The following indicates a quick step-by-step to run the project using Windows PowerShell.
REM 1) Clone the repo (pick one)
git clone git@github.com:KubangPawis/project-sbafn.git
git clone https://github.com/KubangPawis/project-sbafn.git
REM 2) Go into the project folder
cd /d C:\path\to\project-sbafnRun the app (Flutter web):
REM 1.) Navigate to the Flutter app directory
cd sbafn_app
REM 2.) Run the app
flutter run -d chrome --dart-define=MAPTILER_KEY=<INSERT_API_KEY>(Optional) Rebuild data pipeline
To rebuild the data pipeline, follow the set of steps listed below. Large third-party data (e.g., Mapillary imagery) is not stored in this repo due to licensing and size. Use the provided scripts to fetch and prepare data.
REM 1) Create and activate a venv
py -3.11 -m venv .venv
.venv/Scripts/Activate.ps1
REM 2) Install Python dependencies
python -m pip install -U pip
python -m pip install -r requirements.txt
REM 3) Run pipeline
python -m pipeline.coreWarning
Rebuilding the pipeline require large downloads. Downloading all raw feature files can take about ≈24+ hours to complete. Instead, a pre-built features dataset was used to supply the model for the prediction.
Project SBAFN employs a hybrid geo implementation, employing both raster and vector geodata. Specifically through the following metrics:
| Feature Type | Indicator Name |
|---|---|
| Physical Indicators | Curb Inlets |
| Drainage Grates | |
| Vegetation Strips | |
| Open Canals | |
| Curb Height (to add) | |
| Driveway Gaps (to add) | |
| Topographic Data | Absolute Elevation |
| Relative Elevation | |
| Road Slope | |
| Geospatial Data | Road Classification |
| Road Width |
➡️ Full definitions, examples, and edge cases: see docs/INDICATORS.md
- Model: YOLOv11
- Data: Manila street-view corpus (330k+ images); trained on a labeled subset for v1.0
- Output: Physical indicator count per street segment
- Model: LightGBM
- Inputs: Physical indicators, topography, road network, rainfall/reports
- Output: calibrated flood-proneness probability score [0, 1]
➡️ Full model detailings: see docs/MODELS.md
Disclaimer (Pilot): Scores are model-derived estimates and may be inaccurate. Not for emergency use—verify with LGU/DRRM advisories.
Project SBAFN envisions to expand its feature to the following in future versions:
- Scale beyond Manila → Metro Manila → nationwide
- Context-based Story Map generation
- AI Assistant for natural-language queries
- Flood-aware routing (citizen view)
- Data pipeline scheduling for frequent risk update
- Code: Apache-2.0 (see
LICENSE). - Street View Imagery: Mapillary imagery is not included in this repo and is governed by Mapillary’s Terms of Use and attribution rules. We only store metadata and our own annotations.
- Elevation data (DEM): Contains modified Copernicus DEM (GLO-30). © European Union, 1995–2025, produced using Copernicus data and information funded by the European Union — accessed via the AWS Open Data Registry. See Copernicus terms of use and license.
Note
This repository does not grant rights to any third-party datasets. Use of Mapillary/Copernicus materials must follow their respective licenses and terms.
- School: Manuel S. Enverga University Foundation Lucena
- Program: BS Computer Science
- Specialization: Data Science
- Email: vlancealexander@gmail.com
- School: Manuel S. Enverga University Foundation Lucena
- Program: BS Computer Science
- Specialization: Data Science
- Email: petergelodantes@gmail.com
- School: Manuel S. Enverga University Foundation Lucena
- Program: BS Computer Science
- Specialization: Data Science
- Email: rjestuesta@gmail.com
- School: Manuel S. Enverga University Foundation Lucena
- Program: BS Computer Science
- Specialization: Data Science
- Email: juanmacatangga1@gmail.com