Teraki Launches Improved Version of its Efficient, Ai-Compatible Edge Processing on Infineon’s AURIX™ TC3xx Microcontrollers
BERLIN, Oct. 15, 2019 - Teraki has successfully implemented its embedded client for Intelligent Signal Processing on the latest generation of Infineon’s AURIX™ microcontrollers (TC3xx), one of the most commonly used real-time controllers in the automotive industry. A demonstration by both companies shows the ability to simultaneously process 50 signals and reduce the data by 95% at low latency at a consistently low percentage of CPU capacity.
Part of the Teraki product is the DevCenter. It automates continuous training and updating of signal processing, ensuring the models maintain the highest accuracy and efficiency during the full life cycle of a car. In addition, Teraki’s DevCenter lets customers configure their required accuracy levels, train their models and directly measure the results of their own machine learning models.
“We greatly appreciate our ongoing partnership with Infineon and the opportunity it presents to share how Teraki enables OEMs and Tier1s with an automated loop of highly efficient, embedded edge data extraction and training AI-models (cloud and edge) that automatically update to ensure the highest accuracy for results and enable lightweight segmentation,” said Dr. Daniel Richart, CEO and founder of Teraki.
“Teraki’s approach to enable AI functionality addresses one of the fundamental challenges for highly automated vehicles, which is efficient processing of sensor data at the edge,” said Ritesh Tyagi, head of the Infineon Automotive Silicon Valley Innovation Center (SVIC). “Infineon recognizes the value of such technology and looks forward to expanding our cooperation with Teraki to bring this technology combined with Infineon’s AURIX TC3xx microcontrollers to the broader market.”
Customers use the Teraki product to train their AI-models more quickly when in development and to continuously achieve the highest accuracy of their AI-model outcomes, e.g., detection scores, when in production. DevCenter works in real time and without owning the data, whether condensed or fully reconstructed. Both customers’ cloud algorithms and low latency, in-car edge algorithms work with Teraki pre-processed data.
Teraki CTO Markus Kopf stated, “We are very happy with the recent improvements in DevCenter as they give our customers even more control and flexibility of the training and testing of their data. Customers can configure the required accuracy levels, train models in parallel and directly measure the results on their targeted machine learning models. In addition, UX/UI was significantly improved leading to an even better and more intuitive experience when using DevCenter. In the near future we will add a feature allowing a customer to download their optimal model for a directly deployment in the car.”