CUVIS.AI

Build and deploy custom AI workflows for hyperspectral data.

Open Source

Reusable Pipelines

AI Integration

BUILD YOUR HYPERSPECTRAL AI WORKFLOW

RGB images show color, shape, and structure. Hyperspectral data adds information about materials and properties that RGB alone may miss. CUVIS.AI combines this spectral information with processing algorithms and AI models in reusable pipelines for your application.

  • Processing Nodes
    Use building blocks such as channel selection and PCA to prepare hyperspectral data for analysis.
  • AI Models
    Integrate models such as SAM and YOLO, alongside CLIP-based approaches, into your hyperspectral analysis workflows.
  • Connected Pipelines
    Link processing steps through typed inputs and outputs to create a complete analysis workflow.
  • Custom Plugins
    Extend the framework with your own nodes and algorithms to meet specific requirements.
  • Reusable Workflows
    Reuse existing components across projects and add new ones where your application needs them.
  • CUVIS.NEXT Integration
    Bring your pipelines into CUVIS.NEXT to run them on hyperspectral camera data.

REUSABLE BUILDING BLOCKS FOR YOUR APPLICATION

Every application starts with a specific challenge. CUVIS.AI provides 90+ reusable nodes that developers can combine into processing and AI pipelines, adding custom components where needed. Pipelines and plugins can then be reused and adapted across projects.

AI DETECTION WITH SPECTRAL DATA

Foreign objects can resemble the surrounding material in color and shape. CUVIS.AI combines spectral information with AI models to help detect these differences. Developers can build and adapt detection pipelines for their own materials and inspection tasks.

  • Combine spectral processing with AI-based detection
  • Extend pipelines with application-specific models and plugins
  • Run the resulting workflow in CUVIS.NEXT
Foreign object detection in lentils with the ULTRIS XMR. Colored outlines highlight detected anomalies.

AI TRACKING WITH SPECTRAL INFORMATION

Similar-looking objects can be difficult to distinguish in RGB images. CUVIS.AI lets developers combine selected spectral channels with modern vision models in a tracking pipeline. This example shows how additional spectral information helps a tracker follow a person when visual appearance alone is insufficient.

  • Select spectral channels that reveal relevant differences
  • Combine spectral processing with detection and tracking models
  • Compare tracking results using RGB and spectral inputs
Yellow outline: RGB tracking. Green outline: tracking with selected spectral channels.

OPEN SOURCE FOR DEVELOPERS

CUVIS.AI is an open-source framework for computer vision and machine learning with hyperspectral data. Built on PyTorch and PyTorch Lightning, it provides a shared environment for processing, training, evaluation, and deployment.

  • Source Code & Examples
    Explore the framework on GitHub and use existing examples as a starting point for your application.
  • Documentation & Datasets
    Get started with tutorials, workflow documentation, and datasets for development and experimentation.
  • Python & gRPC Access
    Integrate hyperspectral processing and AI pipelines into your own software through Python and gRPC interfaces.
  • Plugins & Extensions
    Add your own algorithms and models through plugins, extending the framework for your specific requirements.

BRING YOUR PIPELINES TO CUVIS.NEXT

CUVIS.AI and CUVIS.NEXT share the same pipeline technology. Developers can build a solution in CUVIS.AI and make it available as a ready-to-use workflow in CUVIS.NEXT. Users can then apply it to recorded measurements or live camera data and view the results without writing code.

Experience CUVIS.NEXT & CUVIS.AI at VISION 2026

CUVIS.NEXT and CUVIS.AI officially launch at VISION 2026. See the software running live, meet our team, and discuss your own hyperspectral application.

  • 6–8 October 2026
  • Messe Stuttgart · Hall 10 · Booth 10D12
BE