Ramzy70/ai-based-crop-yield-classification

Code, data, and documentation for the AI-based crop yield classification project using Sentinel-2 imagery in Algeria. # AI-Based Crop Yield Classification from Satellite Imagery **Enhancing Agricultural Monitoring in Algeria** ## 📌 Project Summary This repository contains the full implementation and data preparation pipeline for our graduation project titled: > **AI-Based Crop Yield Classification from Satellite Imagery: Enhancing Agricultural Monitoring in Algeria** The project proposes an artificial intelligence framework that leverages **Sentinel-2 satellite imagery**, **weather data**, and **field observations** to classify wheat yields in the **Constantine region of Algeria**. It integrates both **machine learning** and **deep learning** models, with a focus on explainability, temporal dynamics, and geospatial consistency. --- ## 📁 Repository Structure - `1- inpv data`: Official data files and yield estimates from INPV for Constantine and Oum El Bouaghi. - `2- ground data acquisition`: Field-level cleaned data, planting maps, and cleaning notebooks. - `3- boundaries QGIS`: Digitized shapefiles and boundary metadata for wheat plots. - `4- diseases`: Field-level disease records with severity assessments. - `5- sentinel 2 data acquisition`: Sentinel-2 data import and management scripts. - `6- data preprocessing for ML`: Feature engineering and metadata augmentation for ML pipeline. - `7- weather data`: Meteorological datasets from NASA and derived indicators. - `8- ML model`: Training notebook, feature importance outputs, and results for the Random Forest model. - `9- data preprocessing for DL`: Scripts for temporal data formatting and Sentinel-2 index generation. - `10- DL model`: PyTorch TemporalCNN architecture, ONNX export, and model performance analysis. --- ## 🧠 Methodology ### ✅ Classical Machine Learning - **Algorithm**: Random Forest - **Features**: Vegetation indices, weather statistics, diseases, agronomic data - **Output**: Yield classification (Low / Medium / High) - **Explainability**: Feature importance analysis with SHAP-like interpretation ### ✅ De …