BantayGasang: Underwater action-cam coral bleaching and crown-of-thorns mapper for Bantay Dagat

Problem Statement

The Philippines sits at the center of the Coral Triangle and has established over 1,800 locally managed Marine Protected Areas (MPAs) and fish sanctuaries to sustain municipal fisherfolk catches. Under the Fisheries Code (RA 8550, as amended by RA 10654), coastal barangays and Bantay Dagat (sea patrol) volunteers guard these sanctuaries. However, monitoring whether a sanctuary is recovering or dying requires certified marine biologists to swim Line Intercept Transects and manually label thousands of underwater quadrat points—costing ₱50,000 to ₱150,000 per survey. Consequently, over 85 percent of barangay fish sanctuaries have no regular ecological monitoring, leaving fisherfolk blind to mass coral bleaching during El Niño marine heatwaves, silent crown-of-thorns (dap-ag) starfish plagues that eat a reef in weeks, and hidden blast/cyanide fishing rubble.

Proposed Solution

BantayGasang turns a 15-minute Bantay Dagat snorkel patrol into a scientific benthic reef survey using any smartphone in a ₱300 waterproof pouch or a ₱2,500 action camera:

1. Snorkel Video Transect Capture: During their weekly patrol, a Bantay Dagat volunteer swims along the sanctuary marker rope at shallow depth (2 to 5 meters) pointing their waterproofed camera downward.
2. Underwater Color Restoration & Keyframe Extraction: Because water absorbs red wavelengths and creates blue-green haze, BantayGasang automatically extracts non-overlapping keyframes every 1.5 meters using visual odometry, filters out blurry frames, and applies Sea-thru / CLAHE red-channel attenuation compensation.
3. Automated CoralNet Point Classification & Threat Detection: On each quadrat frame, a deep-learning classifier modeled on the CoralNet architecture labels 50 stratified random points into standard Philippine reef monitoring categories: Hard Coral (Live), Bleached Coral, Soft Coral, Nutrient Indicator Algae, Abiotic Rubble/Sand, plus a dedicated YOLOv8 object detector that counts Crown-of-Thorns Starfish (Acanthaster planci) and Ghost Nets.
4. Barangay Sanctuary Health Card: Generates a DENR-BMB and BFAR-aligned Reef Health Scorecard (% Live Hard Coral Category: Poor / Fair / Good / Excellent) and maps exact GPS hotpots where volunteers need to conduct manual dap-ag (crown-of-thorns) extraction dives that same afternoon.

Target Market

• Primary users: Bantay Dagat sea patrols, municipal fisherfolk associations, and Barangay Fisheries and Aquatic Resources Management Councils (BFARMCs) across 900+ coastal municipalities.
• Public-sector and NGO buyers: Municipal Agriculture Offices (MAO), BFAR regional offices, DENR Biodiversity Management Bureau (Coastal and Marine Ecosystems Management Program), and marine conservation NGOs (Haribon, WWF-Philippines, Rare, Coastline communities).
• Community benefit: Protects the breeding grounds that directly supply daily food and income to 2 million municipal fisherfolk.

How AI Is Used

• Physics-based underwater image enhancement: Corrects wavelength-dependent light absorption and scattering in turbid coastal water before inference.
• Dense benthic point classification (CoralNet / EfficientNet backbone): Classifies 50 points per frame across 100 transect frames (5,000 points per swim) in under 3 minutes—work that takes a marine biologist two full days at a desk.
• Crown-of-thorns (dap-ag) and blast-rubble object detection: Spots camouflaged thorny starfish arms and freshly shattered branching coral rubble so patrols can intervene before the reef is wiped out.

Community & Ecosystem Integration

• Aligns with UP MSI “Alwan” & DENR Citizen Science Reef Monitoring: Outputs the exact percentage cover metrics used by the University of the Philippines Marine Science Institute and DENR MPA effectiveness awards (MEAT).
• Supports Performance-Based LGU Ecological Fiscal Transfers: Gives coastal municipalities verified reef-health data to unlock national conservation grants and blue-economy funding.
• Offline boat-landing sync: Volunteers process videos at the barangay hall or municipal fish port without needing signal out at sea.

Demo Scenario (3 minutes for judges)

1. Upload a 30-second underwater snorkel video clip from a Philippine reef affected by partial El Niño bleaching and crown-of-thorns starfish.
2. Watch BantayGasang extract 15 non-overlapping quadrat frames, run instant underwater color correction (restoring natural coral hues from blue-green cast), and plot the 50 classification points per frame.
3. Within 10 seconds, the dashboard computes the Sanctuary Scorecard: 38.4% Live Hard Coral (Fair), 19.2% Bleached Coral (Warning), 14.0% Dead Rubble, and flags 4 Crown-of-Thorns Starfish detected at Transect Segment 3.
4. Click the Bantay Dagat Action Brief button to print a one-page Visayan/Tagalog patrol map telling fisherfolk volunteers exactly where to dive with tongs to remove the dap-ag cluster and where to ban boat anchoring during bleaching recovery.

30-Hour Build Plan

Suggested stack: Python FastAPI, PyTorch (timm EfficientNet + Ultralytics YOLOv8), OpenCV, PostgreSQL/PostGIS, Next.js + Leaflet. Public data / APIs: NOAA / UC San Diego CoralNet public Indo-Pacific benthic datasets, CSIRO Great Barrier Reef Crown-of-Thorns Starfish (COTS) Kaggle dataset, NOAA Coral Reef Watch Degree Heating Weeks (DHW) satellite bleaching alert API, PhilReefs public MPA directory.

Research & Literature References

1. Oscar Beijbom, Peter J. Edmunds, David I. Kline, B. Greg Mitchell, and David J. Kriegman, “Automated Annotation of Coral Reef Survey Images,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1170-1177, 2012. DOI: 10.1109/CVPR.2012.6247798
   • How BantayGasang uses this study: Implements the foundational CoralNet multi-scale texture and color patch classification pipeline for automated random-point benthic percent-cover estimation on underwater reef imagery.

Risks & Privacy Notes

• Diver and volunteer safety: Surveys are strictly designed for shallow surface snorkeling (1 to 4 meters) along existing sanctuary marker buoys; volunteers are never asked to free-dive deep walls or handle venomous crown-of-thorns spines without standard extraction tongs.
• Protecting endangered species coordinates: Exact GPS pins of giant clams (taklobo) or high-value commercial species are masked on public views to prevent poaching.
• Strict 30-hour feasibility constraint: Live underwater coral transect capture cannot be demonstrated inside a university hackathon venue, forcing the demo to rely on pre-recorded GoPro reef footage.

AI Use Disclosure (draft)

BantayGasang uses OpenCV for underwater color restoration, an EfficientNet patch classifier modeled on CoralNet for benthic percent-cover estimation, and YOLOv8 for crown-of-thorns starfish detection. AI coding assistants helped build the video keyframe extractor and scorecard dashboard. All ecological health categories follow DENR-BMB and PhilReefs monitoring standards.

Rubric Scoring