Teledyne's New Software Update Boosts Machine Vision

Teledyne DALSA has announced the recent updates of its latest Sapera™ Vision Software Edition. Particularly, the release includes enhanced versions of its AI training graphical tool Astrocyte™ 1.50 and its image processing and AI libraries Sapera Processing 9.50.

Accordingly, Teledyne DALSA’s Sapera Vision Software provides field-proven image acquisition, control, image processing, and artificial intelligence functions. Hence, enabling the design, development, and deployment of high-performance machine vision applications. Among the applications include surface inspection on metal plates, location and identification of hardware parts, plastic sorting, and PCB inspection.

Rotated Object Detection. Image courtesy of Linx Taiwan

“This new version of Sapera Vision Software provides a rich set of new and improved features allowing flexible design and accelerated execution of AI-enabled applications,” said Bruno Menard, Software Director for Teledyne DALSA. In addition, Menard said, “Among the key features of this release the Rotated Object Detection algorithm outputs position and orientation at inference time for more precision when detecting targets and defects.”

New Features, Functions

LINX Taiwan, which provides factory automation tools and Teledyne DALSA client, said, “In the inspection of MicroLED, it is necessary to determine whether the LEDs on the substrate have offset and rotation problems. With Astrocyte tools, it is possible to reduce the program flow and inspection time, and to determine the amount of rotation of the LEDs while judging whether they are good or bad.”

Among the new features in this release allows rotated object detection for precise location and orientation of objects/defects. Moreover, it also has a new segmentation algorithm with higher accuracy and faster training and execution.

In addition, the new updates also provide speed improvement via NVIDIA RTX 4000 series and performance power mode management. It also provides faster training time with the usage of multiple GPUs and mixed-precision quantization.

Meanwhile, another new feature is the inclusion of new series of AI sample applications based on Qt/C++ and .NET/C#. At the same time, it supports RGB packed (24-bit) in AI inference. Finally, the new updates provide ability to acquire live video into sample applications via Teledyne’s Spinnaker acquisition API.