Seco has expanded the functional scope of its Application Hub, which was launched in the summer. The platform provides ready-made, tested AI applications that can run on various edge systems. The aim is to reduce development effort and enable faster integration into industrial products.

The new applications include a locally executable voice model based on the efficient Phi model and optimized with Intel's OpenVINO. This allows chatbots or interactive assistants to be operated directly at the edge without the need for a cloud connection. A speech-to-text solution based on OpenAI's Whisper model, which runs under TensorFlow and enables precise transcriptions - for voice control, accessible applications or meeting protocols, for example - is also intended for speech processing. An OpenVINO version optimized for Intel architectures is in preparation.

For security-critical scenarios, Seco is adding an anti-spoofing algorithm that analyzes videos from standard RGB cameras to distinguish real people from photos or digital images. This is intended to prevent manipulation during facial recognition. Another focus is a time series algorithm that processes sensor data from wearables such as smartbands in real time and supports applications in fitness and health monitoring.

In addition, an image enhancement process is available that increases the resolution and quality of photos and videos through intelligent upscaling. Areas of application range from medical imaging and monitoring systems to the restoration of digital media. The package is rounded off by an application for industrial quality control: it simplifies the training of image analysis models in order to automatically detect production errors and thus reduce rejects and costs.

All applications have been validated by Seco to ensure reliable runtime on its own hardware. The company plans to regularly expand its offering in the coming months.