Teledyne DALSA Astrocyte
No-Code AI training tool to quickly deploy AI models for machine vision solutions


Teledyne DALSA Astrocyte is an AI training tool that quickly deploys AI models with no need for coding or AI expertise. With the Astrocyte specialized algorithms for classification, anomaly detection, object detection, and segmentation tasks and the easy-to-use graphical user interface tools, trained neural networks can be deployed easier, faster, and more affordably.

The easy to use Graphical user interface provides users with the tools to create image samples from live image streams or import image samples, import or create annotations, view, edit, manipulate and export dataset.

Astrocyte AI trained models are well suited for applications such as identifying subtle defects and handling variations in shade and perspective that traditional machine vision algorithms can’t handle.



Features and Benefits of Astrocyte AI training tools


  • No coding or AI expertise is required to create AI models
  • Easy to use Graphical user interface to create image samples from live image streams or import image samples, import or create
  • Annotate, visualize, edit, and manipulate datasets
  • Continual learning to update classification models on the factory floor without retraining
  • Save time with automatic labeling (Via SSOD or pretrained models), or import existing label sets
  • Support for deep learning architectures: classification, anomaly detection, object detection, and segmentation
  • Astrocyte AI specially trained models overcome the obstacles faced by traditional machine vision inspection applications
  • Easily integrate with Teledyne DALSA’s Sapera LT and Sherlock software

Example of traditional machine vision inspection
vs
Astrocyte classification algorithm for inspection

As shown in the example below, Astrocyte AI tools overcome the difficulties that are encountered by traditional machine vision inspection applications.

Traditional machine vision imaging
  • Mistakes water droplets for surface damage
  • Less adaptable to shade and perspective variations
  • Misses subtle surface imperfections
Astrocyte classification algorithm
  • Ignores water droplets
  • Adaptable to shade and perspective variations
  • Detects a wide range of defects and imperfections

illustrates how inspection using Astrocyte classification algorithm performs better than traditional machine vision imaging





Astrocyte Graphical User Interface


With no coding and the easy-to-use graphical user interface user’s can:Train an AI model in under 10 minutes with good data with no coding or AI expertise; Select one of three tuning depth options (low, medium, high) to produce an optimized AI model; Rapidly develop machine vision software solutions

Some examples of the GUI interface are shown below.

Detecting anomolies
Astrocyte annotations dialog
Astrocyte AI training model dialog box

Application Examples


Astrocyte’s AI training tools are an ideal solution for many industrial machine vision and medical imaging applications. Here are some real-world examples of how Astrocyte AI training tools software are currently being used.

Plastic recycling

In plastic recycling Astrocyte’s AI classification algorithm classifies and screens more than 30 types of materials to ensure high accuracy and high-quality material.

Medical Imaging

Astrocyte AI tools have also been used to improve the quality, productivity, and efficiency in X-ray medical imaging. Tiny fibers that can collect on X-ray detectors make it difficult and challenging to identify the defects. Using Astrocyte AI analysis, the defects are identified accurately and more quickly.

Screw Thread Inspection

This example shows how Astrocyte AI tools can be used in inspection applications to classify good and bad screw threads. Small defects can be detected and classified into good and bad parts using just tens of sample images to train a model. Using anomaly detection, a binary classifier (good/bad) trained on “good” images only clearly identifies the defective screw.

An example of bad screw thread detection





Using Astrocyte for Inspection



Integration with Sapera Processing and Sherlock


Easily integrate Astrocyte with Sapera Processing and Sherlock for runtime inference. Combine rule-based algorithms with AI models for a full solution.