The current approach to data collection and crop pest and disease diagnosis is evolving from traditional methods of identifying diseases through visible symptoms to the adoption of data-driven solutions powered by machine learning and computer vision techniques. Smallholder farmers and agricultural experts are increasingly using mobile phones equipped with software to automatically collect geo-coded and time-stamped field-level data.
However, much of the image data previously collected has not been adequately curated, prepared, or shared with the broader machine learning community. Additionally, in many cases, by the time image data is captured, the disease has already spread to different parts of the plant, making intervention difficult. This project focuses on creating open machine learning image and spectral datasets for early detection of crop pests and diseases in key crops like cassava, maize, beans, bananas, pearl millet, and cocoa. Collected from Uganda, Tanzania, Namibia, and Ghana, the data aims to fill existing gaps in curation and sharing. By leveraging mobile technology, machine learning, and computer vision, the project supports timely disease diagnosis—ideally before visible symptoms—empowering farmers with better decision-making tools for improved crop health.