
The toolkit enables the use of deep learning based algorithms for projects such as image recognition, autonomous driving, sensor data fusion, driver monitoring and other automotive applications. The toolkit enables customers to develop applications on desktop, cloud and GPU environments and port neural networks to an eIQ Autocompatible S32 processor. NXP's toolkit and the inference engine specified for automotive electronics make integrating neural networks into applications with high security requirements much easier.
An example of this is the transition from conventional image recognition algorithms to those based on deep learning. The latter promises better accuracy and easier maintenance for object recognition and classification. However, implementation in vehicles has so far been hampered by significantly higher costs and the complexity of the system.
The new toolkit should significantly reduce the effort required for selecting and programming integrated calculation cores for all layers of a deep learning algorithm. For customers, this means faster time to market. The automated selection process increases performance by a factor of 30 compared to other embedded deep learning structures. This performance boost is achieved by making optimum use of existing resources. These benefits enable developers to evaluate, tune, and ultimately realize their applications for maximum performance.