JuliusFx131/Hack-the-Carbon-InstaDeep-Project

"Hack The Carbon" hackathon challenged participants to use ML and open-source satellite imagery to accurately estimate Above-Ground Biomass in African forests. The goal is to generate high-resolution biomass maps that support carbon accounting, conservation, and REDD+ reporting. Submissions were judged on the accuracy, scalability, and clarity. # Model Description **Hack The Carbon 🌍🌱** was an instadeep project that intended to build a machine learning model to estimate forest Above-Ground Biomass (AGB) across Africa efficiently and accurately. African forests are key carbon sinks, but deforestation and degradation can turn them into carbon sources. Using **open-source multispectral satellite imagery** and **ESA CCI biomass data**, the model I built for this project (ranked first and awarded $2500) provided high-resolution (30 m) biomass estimates to support carbon accounting, REDD+ reporting, and conservation policy. # Intended Use The models are designed to: - Enable accurate **carbon monitoring** and reporting. - Support **policy-making and forest conservation** efforts. - Detect **spatio-temporal changes** in biomass. - (Optional) Provide **uncertainty estimates** for policy-grade decisions. This project is ideal for **remote sensing specialists, data scientists, and climate advocates** aiming to track forest health and support climate solutions. # Biomass Inference with Docker This repository provides a Dockerized solution to run biomass inference on satellite image chips using a trained U-Net model. --- ## Project Main Structure ``` ./JuliusFx131/ ├── Dockerfile ├── requirements.txt ├── app.py ├── .dockerignore └── unet_weights.pth # Your trained model file ``` --- ## 1️⃣ Build the Docker Image ```bash docker build --no-cache -t biomass-inference . ``` > `--no-cache` ensures a fresh build. Omit it for faster rebuilds after the first build. --- ## 2️⃣ Run Inference ### Windows (PowerShell) ```powershell docker run --rm -it ` -v "D:\ZINDI\hack-the-carbon\test\chips:/input_chips" ` -v "C:\Users\PC\Downloads\Julius Fx\unet_weights.pth:/app/model_weights.pth:ro" ` -v "D:\predictions:/output" ` juliusfx/biomass-inference:latest ` python app.py ` --chips_dir /input_chips ` --model_weights /app/model_weights.pth ` --output_dir /output ``` ### Linux / macOS (bash) ```bash docker run --rm …