NXP’s car-level deep learning kit improves performance of next-generation automotive applications by 30 times.
Detroit – October 8, 2019 – NXP Semiconductors NV (Nasdaq: NXPI), the world’s largest supplier of automotive semiconductors, has expanded its eIQ machine with the eIQ Auto, a vehicle-level deep learning toolkit. Learn the product line. The toolkit is designed to help customers quickly move from a development environment to an artificial intelligence application integration that meets the automotive industry’s rigorous standards. eIQ Auto can apply deep learning-based algorithms to vision, autonomous driving, sensor fusion, driver monitoring and other evolving automotive applications.
The eIQ Auto toolkit enables customers to develop automotive products in a PC/cloud/GPU environment and integrate their neural network into the S32 processor. NXP’s toolkit and automotive-level inference engine make neural networks easier to deploy in automotive applications with tight security requirements. A good example is the transformation of traditional computer vision algorithms into deep learning-based algorithms in computer vision-based systems.
Deep learning is expected to provide detection accuracy and maintainability superior to “traditional” computer vision algorithms in target detection and classification, but complexity and high cost hinder the integration and implementation of full vehicle regulations.
The eIQ Auto Toolkit helps customers reduce time-to-market by reducing the investment cost required to select and integrate embedded computing engines for each layer of deep learning algorithms. Compared to other embedded deep learning frameworks, the automatic selection process improves the performance of a given model by a factor of 30. The way to achieve this performance is to optimize the use of available resources, reducing time and development effort1. These bonuses help developers evaluate, fine-tune, and deploy applications for maximum overall performance.
Compliance with automotive grade development standards and functional safety requirements is a key advantage of eIQ Auto and S32V integration. eIQ Auto’s inference engine complies with strict development requirements and is compliant with the Automotive SPICE® standard. The S32V processor offers the highest level of functional safety, supports ISO 26262, and meets ASIL-C, IEC 61508 and DO 178 standards.
Kamal Khouri, NXP’s vice president and general manager of advanced driver assistance solutions, said: “The new generation of automotive applications, such as the current integrated driver test vehicle integration solution, is huge, energy efficient and impossible to mass produce. New eIQ toolkit Help our customers deploy powerful neural networks in embedded processor environments with the highest levels of security and reliability.”
NXP’s Auto eIQ Deep Learning Kit and the car-grade S32V chip together provide a powerful foundation for performance, functional safety and quality for next-generation automotive applications. The NXP eIQ Auto toolkit includes:
- Multiple execution options supported by a unified API and running backend selection
- A-SPICE-compliant inference engine that optimizes performance by scheduling tasks on high-efficiency accelerators
- Support for the most advanced CNN/Neural Network
- Optimization layer and optimized network library