In Sama’s latest ebook, CEO Mark Bakker shares the benefits of ML algorithms for pest, disease and weed identification for better crop management, reduced pesticide usage and increased productivity.
Finding high-quality training data for AgTech computer vision models can be challenging and costly—prompting many Machine Learning teams to explore synthetic data as a way to augment or enhance their datasets. Mark will host a roundtable at World Agri-Tech Synthetic Training Data in AgTech: Balancing Pros, Cons, and Best Practices.
Ahead of this, Mark shares a case study of how Sama used computer vision technologies to help start-up company Orbisk on its mission to help restaurants capture data about food waste:
- Orbisk needed a way to ensure its models could be used to accurately identify different foods in all their potential shapes and sizes, from a variety of angles.
- Sama’s solution was to help build and improve the dataset used to train the computer vision algorithms to accurately identify ingredients, including meat, vegetables, fruit, bread, condiments, sauces, and desserts.
- Open feedback loops enabled the rapid escalation and resolution of edge cases and accommodation of new data sources.
- Same accurately labelled hundreds of thousands of food images with a 99% client acceptance rate.
The impact:
- 200,00kg of food diverted from landfills to date
- 70% reduction in food waste for key clients
- Fewer harmful GHG emissions
Read the full ebook here, to learn more about how Sama is tackling agricultural challenges through machine learning.
Join Mark’s roundtable discussion at World Agri-Tech San Francisco on March 19-20, to hear examples from companies who have found the right balance, as well as additional tactics to try instead of—or alongside—synthetic data. Book your place here.