Infineon Technologies expanded its DEEPCRAFT Studio with support for computer vision, adding to its audio, radar, and other time-series data capabilities. This will enable the platform to develop low-power, low-memory edge artificial-intelligence (AI) models for vision across a wide range of applications and use cases. For example, factories can use DEEPCRAFT to implement real-time visual inspection of parts, or shut down machinery in the presence of a human, or in smart-home appliances to monitor objects, humans, or petss.
Aiding the optimization of AI model development for vision-based use cases, the updates to DEEPCRAFT Studio address data collection and preprocessing through model deployment. End-to-end support for vision, along with Infineon’s PSOC Edge high-performance machine-learning MCUs, lets developers add top-of-the-line edge AI features to their products.
The company created a computer-vision demo to highlight state-of-the-art object detection powered by Ultralytics YOLO models, enabling the training of custom computer-vision solutions using DEEPCRAFT Studio. Leveraging Ultralytics YOLO models also significantly reduces training time while maintaining high accuracy.
These additions, along with the platform’s graph-based UX interface, make it possible to iterate on models during the development cycle. Models can run in real-time using almost any given computer and camera, ensuring it works as intended before deploying onto hardware. This accelerates development, as improvements can be made to the model prior to deployment.
