Food safety ML for pathogen detection with explainable AI
Innovative machine learning solution combining image and handcrafted features to enhance pathogen detection in food, while improving model explainability and managing uncertainty. It aims to minimize false positives, enhance decision transparency, and optimize sample collection through active learning.
Overview
Food safety ML is a cutting-edge machine learning solution designed to enhance pathogen detection in food by combining latent features from images with handcrafted features. This innovative approach aims to create robust models that improve detection accuracy while reducing false positives. The solution prioritizes model explainability, allowing users to understand the decision-making process by identifying key image pixels or features involved. By effectively managing uncertainty and strategically collecting data samples, it balances accuracy, cost-effectiveness, and safety in food safety decision-making.
Technical specifications
Key features:
- Utilizes latent image features and handcrafted features for robust model creation
- Improves model explainability by identifying key features used in decision-making
- Predicts and manages uncertainty to optimize decision outcomes
- Active learning techniques for strategic sample collection, focusing on uncertain instances
- Plans to implement compositional autoencoders for encoding multimodal data and semi-supervised learning for feature extraction
Technology readiness level
This solution is currently at Technology Readiness Level 5, indicating that it has been validated in relevant environments but requires further development and validation for commercial deployment.
About Iowa State University
Iowa State University is a large, comprehensive public land‑grant research university based in Ames, known for combining fundamental discovery with translational, industry‑relevant work. Companies collaborate through a research and technology park that co‑locates corporate R&D with faculty labs and startups, creating steady talent pipelines. Pilot‑scale facilities, field test sites, and a statewide extension network support prototyping, validation, and deployment with partners across the region. Research is supported by competitive federal funding, including awards from NSF, USDA, and DOE. A dedicated technology transfer office and affiliated research foundation streamline IP, licensing, and startup formation, with incubator space on site.
delivered to your inbox.