04
AI/ML & Automation·Case 04·2024

Cassava Disease Detection at Scale

ML PipelineEdge DeploymentAgricultural Systems
Industry

Tunde Fashola Industries

Sponsor & Co-designer

01
The Problem

Plain language

Cassava blight and mosaic disease destroy an estimated 30% of Nigeria's cassava yield annually — equivalent to $3.5 billion in losses. Smallholder farmers in rural areas lack access to agronomists. By the time disease is visible to the naked eye, it has already spread to neighbouring plots.

02
Methodology

SE approach applied

Designed an end-to-end ML pipeline: image dataset curation (12,000 annotated leaf images), CNN architecture selection and fine-tuning for low-compute mobile inference, human-machine interface design for feature phone compatibility, and deployment systems architecture for offline-first operation.

Disciplines Applied

ML PipelineEdge DeploymentAgricultural Systems
Key Outcomes
87%
Disease detection accuracy
<3MB
Model size — offline-ready
14,000
Farmers registered in 6 months
03
Outcome

Impact & results

87% disease detection accuracy with a model under 3MB — operable on ₦12,000 smartphones. Deployed across 3 states through partnership with Borno State Agricultural Development Programme. 14,000 farmers registered in the first six months.

TF

Tunde Fashola Industries

Sponsor & Co-designer

Industry2024

Published by

ASES · UNILAG

Association of Systems Engineering Students