NXP Semiconductors announced the launch of eIQ Auto, a deep learning toolkit for automotive applications, expanding its eIQ machine learning product portfolio. The toolkit aims to enable customers to rapidly transition from development environments to deployment of AI applications that meet stringent automotive standards.

eIQ Auto supports the utilization of deep learning-based algorithms for vision, driver replacement, sensor fusion, driver monitoring, and other evolving automotive applications.

Using the eIQ Auto toolkit, customers can conduct automotive production development in desktop/cloud/GPU environments and deploy neural networks on S32-supported processors.

NXP's toolkit and automotive inference engine enable easy deployment of neural networks in applications with stringent safety requirements. A good example is accelerating the transition from conventional computer vision algorithms to deep learning-based algorithms for vision-enabled systems.

eIQ Auto Block Diagram

Deep learning guarantees enhanced accuracy in object detection and classification and improved maintainability compared to "conventional" computer vision algorithms, but the barriers to complete automotive implementation create significantly increased complexity and cost burden.

The eIQ Auto toolkit aims to reduce customers' time-to-market by lowering the investment costs required to select and program embedded computing engines for each layer of deep learning algorithms.

Through automated selection processes, for certain models, the toolkit delivers performance enhanced 30x compared to other embedded deep learning frameworks. This performance improvement is achieved by optimizing utilization of available resources and reducing time and development effort. As a result, developers can evaluate, fine-tune, and deploy their applications to maximize overall performance.

Compliance with automotive development standards and functional safety requirements is a key benefit obtained through eIQ Auto and S32V integration. The eIQ Auto inference engine has been developed to meet stringent requirements and complies with Automotive SPICE®. The S32V processor provides the highest level of functional safety, supporting ISO 26262 to ASIL-C, IEC 61508, and DO 178.

Kamal Khouri, Vice President and General Manager of NXP's Advanced Driver Assistance Division, stated, "Next-generation automotive applications, as seen in current autonomous vehicle test implementations, are large, power-consuming, and lack practicality for application in mass automotive production. The newly introduced eIQ toolkit enables customers to deploy powerful neural networks in embedded processor environments with optimal safety and reliability."