🌍 GaiaNet — Biodiversity Intelligence Platform (Hackathon Edition)

AI-Powered Ecosystem Monitoring using Google Gemini 2.5 Pro

---

Overview

GaiaNet is an intelligent biodiversity monitoring and conservation decision-support system built for the SEED Hackathon 2025.
It uses Google Gemini 2.5 Pro, Streamlit, and smart ecological modeling to generate real-time insights about wildlife and ecosystems.

🔎 What GaiaNet Does

• Detects wildlife species from images
• Forecasts population decline or recovery
• Models ecosystem stability and species interactions
• Generates conservation actions prioritized by impact & urgency
• Provides a clean, modern dashboard UI with dark mode

This project showcases how multimodal AI can transform conservation and ecological research when combined with structured workflows and intuitive visualization.

---

🏗️ Tech Stack

Component	Technology
Frontend / Dashboard	Streamlit (Dark Mode Custom UI)
AI Engine	Google Gemini 2.5 Pro
Fallback Model	CLIP zero-shot (HuggingFace Transformers)
Data Handling	pandas, matplotlib
PDF Export	fpdf2
Audio Processing	librosa, soundfile
Map Visualization	pydeck
Satellite/Drone Analysis	numpy (mock NDVI)

---

    ┌──────────────────────┐
    │  Streamlit Dashboard │
    │  (app.py - Gemini)   │
    └───────────┬──────────┘
                │
                ▼
    ┌──────────────────────┐
    │ Google BigQuery      │
    │ gaia_net_dwc_data    │
    │ ├ gbif_occurrences   │
    │ ├ wdpa_temp_1        │
    │ └ iucn_species       │
    └───────────┬──────────┘
                │
    ┌─────────────────────────────┐
    │ Google Cloud Run Ingestor   │
    │ (gaianet-ingestor service)  │
    │ /ingest/gbif   /wdpa /iucn  │
    └───────────┬────────┬────────┘
                │         │
    ┌───────────┘         └──────────────┐
    │                                     │
    ▼                                     ▼

GBIF API WDPA + IUCN APIs (occurrence DWCA) (status, PA data)

--

🚀 Key Features (Current Version)

🦜 1. Species Detection

• Upload an image
• Gemini identifies:
  • common name
  • scientific name
  • habitat type
  • confidence score
  • observations
• Automatic fallback to CLIP if API is missing

---

📈 2. Population Forecasting

• Upload or auto-generate population history
• Gemini uses ecological reasoning to produce:
  • next 6 months population forecast
  • confidence intervals
  • decline/recovery assessment
  • textual explanations
• Clean visualization of forecasted trends

---

🕸️ 3. Ecosystem Interaction Modeling

With reasoning:

• Keystone species
• Predation & competition
• Network stability score
• Collapse risk
• Simulation: “what if species declines by 30%”

---

🌱 4. Conservation Recommendations

• Gemini suggests prioritized actions:
  • habitat restoration
  • anti-poaching
  • invasive species control
  • corridor rebuilding
• Provides impact scores, urgency levels, and rationale

---

🧪 5. Clean Dark-Themed Dashboard

• Fully redesigned UI
• Card layout
• Modern typography
• Compact charts
• Clean tab structure
• Consistent dark mode

---

Cloud Run Ingestor

You have deployed: gaianet-ingestor Endpoints:

Endpoint Description

/ Health check /ingest/gbif Starts GBIF DWCA download /ingest/wdpa (Planned) Load WDPA CSV/Geo data /ingest/iucn (Planned) Load conservation status

Credentials are stored as Cloud Run environment variables:

GBIF_USERNAME GBIF_PASSWORD GBIF_EMAIL

⚙️ Installation & Run

1. Install Requirements

pip install -r requirements.txt

**2. Set Gemini API Key

You can provide the key in one of two ways:

Via environment variable

```bash
export GEMINI_API_KEY="YOUR_KEY"

Or enter it directly in the Streamlit sidebar

The sidebar includes a secure password field where you can manually enter your Gemini API key.

---

3. Run the App

streamlit run app.py

🚧 WORK IN PROGRESS — Upcoming Features

GaiaNet is actively being expanded.
The following features are in development and partially implemented in prototype form:

---

📊 1. Beautiful KPI Cards

Upcoming visually appealing KPI-style metrics:

• Ecosystem health score
• Population risk score
• Habitat quality metric
• NDVI (vegetation index)

These will appear after species detection, forecasting, or ecosystem modeling.

---

📄 2. PDF Report Export

A one-click Download Report button will generate a full conservation report containing:

• Detected species
• Forecast graphs
• Ecological reasoning
• Intervention recommendations
• Metadata
• Images / spectrograms

Built using fpdf2 for lightweight PDF creation.

---

🖼️ 3. Multi-Image Batch Detection

Allows users to:

• Upload 5–20 images at once
• Run species detection on each image
• Display results in a gallery-style layout
• Optionally extract & display GPS coordinates (EXIF) on a map

---

🎧 4. Audio Species Detection

Support for analyzing wildlife audio such as:

• Bird calls
• Frog croaks
• Mammal sounds
• Insect signals

Pipeline:

1. Convert audio → spectrogram
2. Send spectrogram → Gemini for species inference
3. (Optional) Use YAMNet for pre-filtering the sound category

---

🛰️ 5. Satellite Image Analysis

Two upcoming satellite/drone modules:

• NDVI vegetation health estimation (mock or real NDVI)
• Habitat quality scoring
• Deforestation / disturbance analysis

Will support drone imagery and geospatial raster data.

---

🗺️ 6. Map Visualizations (pydeck)

A map dashboard layer is in development to:

• Plot species sightings on a map
• Cluster multiple detections
• Visualize ecosystem risk by region
• Use GPS EXIF data when available
• Fallback to synthetic coordinates for demo images

---

🧭 7. Literature Extraction (Planned)

Using Gemini to extract structured ecological insights from research papers:

• Species references
• Threats
• Policy recommendations
• Geographic mentions

This will help tie scientific literature to real-time species detection and forecasting.

Data Acknowledgment

IUCN Red List

This project uses species conservation status data from:

IUCN 2025. The IUCN Red List of Threatened Species. Version 2025-2.
Available at: https://www.iucnredlist.org

All rights reserved by IUCN.

**UNEP-WCMC and IUCN (2025), Protected Planet: The World Database on Protected Areas (WDPA) [Online], November 2025, Cambridge, UK: UNEP-WCMC and IUCN. Available at: www.protectedplanet.net.

---