Actively learned materials segmentation and classification software for scrap analysis
This software utilizes active learning and automated image segmentation to classify bulk scrap materials, overcoming data labeling challenges. It provides comprehensive material analysis and can be customized for various industrial applications.
Overview
Geometric Data Analytics has developed an innovative software solution for the segmentation and classification of bulk scrap materials using active learning techniques. This system addresses the challenges of heterogeneous digital images in scrap analysis by efficiently utilizing expert labelers to improve classification accuracy. The software is designed to overcome the limitations of current generative AI methodologies by providing comprehensive statistical summaries and rigorous machine error values, ultimately enhancing material analysis.
Technical specifications
Key features:
- Combines active learning with automated digital image segmentation
- Utilizes expert labelers to assign categories to ambiguous materials, improving classifier accuracy
- Offers comprehensive statistical summaries of material properties and geometry
- Adaptable to various industrial contexts, leveraging previous successes in plant/soil sciences and wildfire fuels analysis
- Supports a hierarchical labeling scheme and active learning for continuous model improvement
Technology readiness level
This software has reached Technology Readiness Level 6, indicating that it has been demonstrated in a relevant environment. Future validation will involve customization and deployment in industrial settings, focusing on establishing a hierarchical labeling scheme, producing a baseline classification model, and customizing material analysis based on specific needs.
About Geometric Data Analytics
Geometric Data Analytics (GDA) is a Durham, North Carolina-based research, development, and consulting company that specializes in solving complex data analysis problems. The firm is built upon expertise in topological data analysis, applied mathematics, machine learning, and software engineering. Their team develops custom algorithms and software architectures, often working in domains where standard off-the-shelf artificial intelligence and machine learning solutions are insufficient. By utilizing test-driven development and modern, scalable microservice architectures, GDA provides interoperable and maintainable technical solutions designed for deployment across diverse environments, including cloud infrastructures and secure, isolated systems.
The company serves clients in the government, military, and commercial sectors, offering capabilities in areas such as anomaly detection, high-dimensional data analysis, agent-based modeling, and signal processing. GDA focuses on delivering scientific research and algorithmic development that can be seamlessly integrated into larger systems. Their methodology emphasizes speed and reliability, enabling partners to progress from theoretical concepts to functional, deployable prototypes efficiently. GDA also supports open-source initiatives and provides consulting to help organizations modernize their development pipelines through CI/CD practices and containerized deployment technologies.
delivered to your inbox.