ML Capstone Projects

Crop Health Stress Mapping

Synthetic-data capstone

An end-to-end crop-stress mapping capstone using synthetic drone vegetation-index data, with a leakage audit, spatial validation and interpretable maps.

Capabilities

What it does

Trains classifiers to map Healthy and Stressed observations from multispectral and canopy features.
Compares baseline and tuned models with stratified cross-validation, feature selection and threshold calibration.
Audits label leakage, tests spatial-block generalisation, and reports SHAP and permutation importance.
Aggregates cell-level predictions into severity maps and an inspection-plan workflow with a local prediction API.

Compatibility

Platforms and requirements

Python 3.13+ · scikit-learn · SHAP · FastAPI

Synthetic multispectral dataset included in the repository
Python 3.13 or newer

scikit-learn
Geospatial
SHAP
Remote Sensing

Setup

Install and run

Installshell

python3.13 -m venv .venv
pip install -r requirements.txt

First steps

Train the pipeline

Exampleshell

python -m src.train
uvicorn app.main:app --port 8000

Browse ML Capstone Projects