Dragonfly 3D World is a GPU-accelerated volumetric image analysis platform developed by Comet Technologies for industrial computed tomography, non-destructive testing, materials science, and research imaging. It combines traditional image processing with deep-learning-based segmentation—including Meta's Segment Anything Model (SAM)—enabling users to isolate defects, phases, and regions of interest in 3D datasets with far fewer manual operations than thresholding-based workflows require. The software handles datasets from virtually any CT, X-ray microscopy, or synchrotron tomography system.

How does AI-powered segmentation speed up CT data analysis?

Dragonfly integrates convolutional neural network architectures—U-Net, U-Net SR, and Sensor3D—built on TensorFlow/Keras, alongside Meta SAM for one-click intelligent segmentation. Users train custom deep learning models directly within the software using an interactive visual feedback interface that displays training progress in real time. Once trained, a model segments thousands of slices in seconds, replacing hours of manual painting and threshold adjustment. Traditional methods—region growing, watershed, clustering, and multi-threshold Otsu—remain available for datasets where deep learning is unnecessary or the user prefers classical approaches.

What quantitative analysis can Dragonfly perform on CT volumes?

The platform provides sub-voxel precision surface determination for metrology-grade measurements: wall thickness analysis with color-mapped deviation overlays, porosity characterization (size, sphericity, connectivity, spatial distribution), fiber orientation and length distribution in composites, particle size and shape analysis, and CAD-to-scan deviation comparison for dimensional conformance. Results export as CSV, images, or 3D surface meshes in STL and PLY formats. For multi-phase materials, individual phases can be isolated, quantified, and rendered independently.

How does Dragonfly support high-throughput or automated inspection?

Every operation in Dragonfly can be recorded as a macro and replayed across an unlimited number of datasets through the batch processing engine. For more complex workflows, a built-in Python console provides full access to the Dragonfly API, enabling custom analysis pipelines that integrate with open-source scientific Python libraries. The optional Dragonfly Compute Framework (DCF) distributes processing across cluster nodes for large-scale studies involving thousands of CT volumes.

Frequently Asked Questions

What CT systems is Dragonfly compatible with?

Dragonfly imports data from all major industrial and scientific CT manufacturers—Comet Yxlon, ZEISS, TESCAN, Rigaku, Bruker, Nikon, and GE—via standard formats including TIFF stacks, DICOM, BMP, PNG, and MRC. No proprietary file format lock-in; any CT system that exports image stacks can feed data into Dragonfly.

Is Dragonfly available as a network deployment?

Yes. Dragonfly VizServer provides centralized application deployment for enterprise environments, allowing multiple users to access Dragonfly sessions from standard workstations or thin clients while processing runs on GPU-equipped servers.

What training and support are included?

Comet Technologies provides the first year of software maintenance at no charge, including version upgrades and priority technical support. Training resources include video tutorials, application-specific webinars, and an online knowledge base. On-site and remote training packages are available for team onboarding.