AI vision chip for autonomous driving reaches tape-out stage
The tape-out of the AI chip was achieved together with the Fraunhofer Institute for Integrated Circuits IIS. It is an important milestone of the joint research project TEMPO (Technology and hardware for neuromorphic computing) on the way to strengthening the European semiconductor supply chain. Together with automotive application partners such as Valeo and InnoSenT, the AI chip will be used in the TEMPO project to accelerate deep learning inference for automotive use cases, on a small silicon footprint and with low power consumption. Currently, such use cases require much larger high-end computing units, which also require significantly more power.
At the heart of the AI chip is Videantis’ v-MP6000UDX unified processing platform, which will run various neural networks to process radar, lidar or camera data for object detection, classification and localization. It is a highly scalable multicore architecture combined with a custom bus fabric and shared on-chip SRAM with multiple banks. This architecture is expected to provide exceptionally high efficiency and performance for a variety of algorithm types such as deep learning, computer vision, image processing, and video coding. The project also makes extensive use of Videantis’ comprehensive v CNNDesigner front-to-end toolflow, which enables automatic mapping and optimization of neural networks to the target architecture without requiring deep programming knowledge.
Videantis GmbH (Hannover, Germany) is a supplier of deep learning and computer vision solutions based on its unified processing platform. With its processor IP, hardware/software-based solutions for deep learning, computer vision, image processing and video coding, Videantis globally supports semiconductor manufacturers, automotive OEMs and tier 1 suppliers.
TEMPO is a European innovation project. It has received funding from the ECSEL Joint Undertaking (JU). The JU receives support from the European Union’s Horizon 2020 research and innovation programme as well as support from Belgium, France, Germany, Netherlands and Switzerland. Consortium members include Bosch Imec, CEA-LETI, various Fraunhofer entities, Infineon, STMicroelectronics, Valeo, and Videantis.
Related articles:
Continental, videantis join forces for AI-based cameras
Videantis claims 1000x performance boost on deep learning applications
Video processing platform reduces costs, makes systems fail-safe
19 European partners rally on next-gen neuromorphic memory
