CropMD

AI crop disease diagnosis on a basic phone, and a national picture of crop health for the institutions responsible for it.

What it does

A farmer points a phone camera at a leaf or a pod and receives a diagnosis in under a second, along with a treatment plan, an estimate of the financial impact, and the option to ask follow up questions by voice in their own language. None of this requires an internet connection, because the diagnostic models run on the phone itself. Every diagnosis also contributes, anonymously and at district level, to a national surveillance dashboard that gives agricultural authorities a current picture of disease across the country.

Who uses it

Smallholder farmers working one to five acre plots on basic Android handsets are the primary users. Extension officers, cooperative leaders and NGO field agents use it as a diagnostic aid in the field. Institutional users, including national agricultural bodies, regulators and researchers, use the surveillance dashboard for outbreak detection, resource allocation and planning.

Capabilities

Multi-Crop Disease Diagnosis

Diagnosis across 45 disease classes covering seven crops: cocoa, cashew, rubber, oil palm, cassava, maize and tomato.

1/8

Technical approach and regulatory alignment.

The technical approach

CropMD uses a tiered architecture. Small quantised models running on the handset handle the majority of diagnoses instantly and at no marginal cost. Cases where the on device model is not confident, or where the disease is rare, are routed to a cloud vision model that produces a fuller assessment. A conversational layer answers follow up questions with the full context of the diagnosis, location, season and scan history.

The system improves continuously. Farmer corrections become training labels, low confidence images are relabelled by an expert model, and retraining runs automatically once enough new verified images have accumulated for a crop, with a new model promoted only when it outperforms the one in production.

Institutional and regulatory alignment

Methodology aligned with national agricultural and cocoa sector authorities, tree crop and rubber sector bodies, and national research institutes for variety catalogues and disease taxonomy. Treatment advice is integrated with the national approved pesticide registry. The platform complies with Ghana's Data Protection Act, Act 843.