Case study · AI in production
AI Seed Quality Analysis and Sorting Platform
A cloud platform and sorting technology that uses computer vision and deep learning to judge the quality and genetic traits of individual seeds from images, without destroying them. Our engineers work on the product's software.
AIThe client
Seed companies need to know which seeds will germinate, grow vigorously and stay genetically true, and traditional tests are slow and destroy the seed. This platform images every seed and predicts traits such as germination, vigour, uniformity and genetic purity with no DNA extraction. Seed companies use it for research and breeding, and AI-enabled sorting machines use it to upgrade seed lots.
The challenge
- Turn thousands of data points from each seed image into reliable quality predictions.
- Deliver results to researchers in minutes, through a secure cloud platform.
- Keep each customer's data separate while letting models improve from shared, anonymised images.
- Feed trained classifiers into industrial sorting machines on the production line.
What we did
Our engineers work within the client's product team on the software around the AI: the cloud platform that receives and processes seed images, the analysis results customers see, and the integration that carries trained classifiers to sorting machines. The client's scientists own the models; WeyBee's engineers build and maintain the product that puts them to work.
What it does
Seed-by-seed analysis
Each seed imaged and assessed individually, without destroying it.
Trait prediction
Germination, vigour, uniformity and genetic purity predicted from appearance.
Cloud results platform
Results in minutes as histograms, similarity matrices and scatter plots.
AI-powered sorting
Trained classifiers run on optical sorting machines to upgrade seed lots.
Research reports
Morphology and quality correlation reports for breeding programmes.
Data protection
Customers see only their own data, with dedicated cloud options.
Results
- Used by more than 50 seed companies across around 30 crops.
- Low-grade seed lots upgraded to market value through AI-powered sorting.
- Researchers get seed-level data in minutes instead of waiting for grow-out tests.
Computer vision that grades
seeds one by one
The client's platform uses computer vision and deep learning to judge the quality and traits of individual seeds from images, without destroying them. Our engineers work on the software around the AI.
The cloud platform receives and processes seed images from labs and sorting machines.
Analysis results and research reports presented for seed companies and breeders.
Machines use the analysis to upgrade low-grade seed lots to market value.
Customer data is protected across the platform and its integrations.