Denso and Toshiba have reached a basic agreement to jointly develop an artificial intelligence technology to be used in image recognition systems independently developed by the two companies for advanced driver assistance and automated driving technologies.
The technology – called Deep Neural Network-Intellectual Property (DNN-IP) – is an algorithm modeled after the neural networks of the human brain, and is expected by Denso and Toshiba to perform recognition processing as accurately as, or even better than the human brain.
To enable automated driving, automotive computers need to be able to identify different road traffic situations including a variety of obstacles and road markings, availability of road space for driving, and potentially dangerous situations. In image recognition based on conventional pattern recognition and machine learning, objects that need to be recognized by computers must be characterized and extracted in advance.
In DNN-based image recognition, computers can extract and learn the characteristics of objects on their own, thus significantly improving the accuracy of detection and identification of a wide range of objects.
Because of the rapid progress in DNN technology, the two companies plan to make the technology flexibly extendable to various network configurations. They will also make the technology able to be implemented on in-vehicle processors that are smaller, consume less power, and feature other optimizations.
Denso has been developing DNN-IP for in-vehicle applications. By accelerating the process to commercialize DNN-IP through the joint development and incorporating DNN-IP in in-vehicle cameras, Denso says it will “develop high-performance, advanced driver assistance and automated driving systems, and continue to contribute to building a safe and secure automotive society for people around the world, not just for drivers and pedestrians.”
In addition to its conventional image processing technologies, Toshiba will partition this jointly developed DNN-IP technology into dedicated hardware components and implement them on its in-vehicle image recognition processors to improve their image processing performance and enable them to process images using less power than image processing systems with digital signal processors (DSPs) or graphics processing units (GPUs).