Here is the complete, perfectly formatted "About the Project" submission for Devpost, formatted with GitHub-flavored Markdown and LaTeX math support:

🌿 Re-Grow: Active Forest Disaster Recovery Platform

💡 Inspiration

Over 10 million hectares of forest are lost globally every year to illegal deforestation and catastrophic wildfires. While modern Earth Observation satellites capture high-resolution imagery of these environmental disasters, there remains a massive gap between disaster detection and ground-level ecological recovery.

Environmental response teams often spend weeks manually measuring damage footprints, calculating sapling inventories, estimating recovery costs, and searching for local conservation organizations to execute planting campaigns. We built Re-Grow to eliminate this friction — transforming raw satellite imagery into actionable, real-time reforestation blueprints in under 3 seconds.

🌱 What It Does

Re-Grow is an end-to-end satellite Earth Observation platform that bridges computer vision analytics with real-world ecological restoration:

📡 Multi-Source Satellite Image Acquisition: Users can upload custom drone/aerial imagery or stream live, cloud-free Sentinel-2 STAC satellite tiles for any global location (Amazon, California, Black Forest, Maui, etc.).
🔬 Dual Computer Vision Engine:
- Deforestation Analysis: Detects tree cover loss, clear-cutting scars, and barren soil conversions using multi-spectral greenness ratio variance and intensity change detection.
- Forest Fire Detection: Pinpoints active fire hotspots and burn scar boundaries.
📊 Automated Reforestation Economics: Quantifies total damaged area ((m^2)), computes the required sapling inventory using forestry density standards ((1 \text{ tree per } 4m^2)), and calculates total estimated USD recovery budgets.
🔥 Live Climate & Fire Risk Telemetry: Integrates real-time weather data (temperature, wind speed, wind heading) to compute active wildfire propagation risk scores.
🌐 50km Geospatial NGO Matchmaker: Queries the OpenStreetMap Overpass API to locate registered forestry offices, environmental non-profits, and nature clubs within a 50km radius for rapid deployment.

📐 Mathematical Framework

The economic recovery engine relies on quantitative forestry formulas:

1. Required Sapling Count ($S$)

Given a detected damaged area $A$ (in $m^2$) and a target planting density $\rho = 0.25 \text{ trees}/m^2$ ($1 \text{ tree per } 4m^2$): $$S = \lceil A \times \rho \rceil$$

2. Total Estimated Budget ($C_{\text{total}}$)

With a benchmark purchase, soil preparation, and labor cost $P_{\text{sapling}} = \$2.50 \text{ USD}$: $$C_{\text{total}} = S \times P_{\text{sapling}}$$

3. Vegetation Index Variance ($\Delta V$)

Multi-spectral greenness decay across baseline ($I_{\text{before}}$) and post-disaster ($I_{\text{after}}$) satellite bands: $$\Delta V = \max\left(0, \frac{G_{\text{before}}}{\bar{R}{\text{before},\text{blue}} + \epsilon} - \frac{G{\text{after}}}{\bar{R}_{\text{after},\text{blue}} + \epsilon}\right)$$

⚙️ How We Built It

Frontend: Built with Next.js 15, React, TypeScript, and Vanilla CSS featuring interactive before/after split comparison sliders, multi-mode view tabs, glassmorphism UI, and OpenStreetMap Leaflet mapping.
Backend API: Engineered a high-performance Python FastAPI REST server powering the computer vision algorithms, database audit logs, and external API integrations.
Computer Vision & Geospatial: Powered by OpenCV, scikit-image, NumPy, and Pillow for adaptive thresholding, morphological noise reduction, and region prop analysis.
Data Integrations:
- Microsoft Planetary Computer STAC API & ArcGIS World Imagery for live cloud-free satellite Earth Observation tiles.
- OpenStreetMap Overpass API for 50km radius geospatial NGO querying.
- Open-Meteo API for live climate telemetry.

🔥 Challenges We Faced

Handling Varied Satellite Tile Formats: Real-world satellite imagery varies drastically in cloud cover, sun angles, and color channels. We developed a combined multi-spectral and intensity change detection algorithm to ensure robust damage quantification across any geography.
Real-Time Satellite Streaming: Direct STAC API queries can sometimes be constrained by asset authentication or cloud cover. We implemented a multi-tier failover strategy (STAC Search (\rightarrow) ArcGIS Earth Observation Tile Export (\rightarrow) Real High-Res Satellite Datasets) guaranteeing zero downtime.
Cross-Origin Vercel Deployment: Proxying REST API endpoints from a Vercel-hosted frontend to a Python FastAPI backend required custom Next.js environment rewrites to eliminate HTTPS mixed-content blocks.

🧠 What We Learned

How to query and process STAC (SpatioTemporal Asset Catalog) payloads for real-time satellite Earth observation.
How to apply morphological operations (erosion/dilation) on satellite arrays to filter out noise while preserving distinct damage clusters.
Designing responsive, fluid user interfaces that turn complex geospatial data into accessible, decision-ready visual insights.

🚀 What's Next for Re-Grow

Autonomous Drone Fleet Integration: Exporting calculated damage coordinate bounds to autonomous seed-dropping drone fleets for automated aerial reforestation.
SAR Radar Penetration: Integrating Synthetic Aperture Radar (Sentinel-1 SAR) to penetrate dense cloud cover in tropical rainforests during active monsoons.
Blockchain Carbon Credit Verification: Issuing verified carbon credit certificates to partner NGOs based on post-planting satellite growth tracking.