The AI Box is a powerful GPU workstation with 512 CUDA and 64 Tensor Cores, designed for fast and efficient evaluation of images with Artificial Intelligence (AI).

This edge device was developed for use in a rough industrial environment - fanless, energy-efficient and designed for a wide temperature range. Power can be supplied via PoE+ or via a separate 24V supply. As a field device, in an IP67 housing, it meets the high requirements for shock and vibration resistance. All connections are easily accessible and designed as robust M12 round plugs.

The design as a separate module makes it possible to upgrade older systems with artificial intelligence. This allows you to remain flexible and to continue using your existing application.

Via the integrated web server, previously trained AI models can be loaded onto the AI Box and configurations can be done.

The image processing software communicates with the AI Box via the REST interface. Requests, including an image, are transmitted via http protocol. Only a few tenths of a second later, the evaluation results of the AI are available and can be converted and output via the machine interface.

The necessary training of the AI algorithms can be done independently on a separate computer or with the VisionCockpit cloud application. Depending on the task, extensive pre-trained neural networks are available for selection.