- Software
- Analog Expert Software and Tools
- Automotive Software and Tools
- Functional Safety Software
- DSC Developer Resources
- i.MX Developer Resources
- Developer Resources
- QorIQ Developer Resources
- RFID Developer Resources
- Sensor Toolbox
- Development Software
- Embedded Software
- eIQ Auto ML SW Kit
- eIQ AI Development Environment
- Software Library
eIQ® Auto ML Software Development Kit
NXP eIQ® Auto Machine Learning (ML) software development kit empowers developers to build intelligent automotive solutions—without requiring deep AI expertise. This comprehensive development suite enables seamless preparation, compilation and deployment of ML applications across NXP' s S32 automotive processors and microcontrollers, including Arm® Cortex® cores and Neural Accelerators. With its flexible, scalable framework and open ecosystem, eIQ Auto simplifies the integration of real-time intelligence into vehicles while maintaining core values of accuracy, performance, safety, efficiency and trust.
eIQ® Auto Development Flow
eIQ® Auto Machine Learning (ML) Toolkit
- Comprehensive framework to optimize and deploy ML on S32 automotive microcontrollers and microprocessors
- Extensive optimization tools and Model Zoo
- Unified APIs with multiple choices of inference engines, including an automotive quality option
- Run ML workloads on multi-cores
eIQ® Auto Example Use Cases
NXP eIQ® Auto Machine Learning (ML) software development kit flexibility allow the user to deploy a wide range of ML algorithms, to achieve the compute resources needed in a SDV E/E architecture. By selecting the appropriate automotive microcontrollers and microprocessors, this unified ML software development kit addresses the next use cases:
| Automotive Example Use Cases | Product | Accelerator(s) |
|---|---|---|
| Intelligent Data Orchestrator Intelligently manage AI/ML and other tasks through your vehicle ensuring deterministic task execution |
S32G | Arm® A53 (Neon) |
| Virtual Sensors Replace hardware sensors or deploy in previously inaccessible places like inside motors |
S32 Automotive Processing Platform | eIQ® Neutron (NPU) Arm® R52 (Neon) Arm® A53 (Neon) DSP AI/ML Accelerator Arm® M7 |
| Predictive Maintenance Utilize real-time data from individual vehicles to prevent anomalies in your fleet |
||
| Imaging Radar Radar-based object classification with reduced dependencies on cameras or LiDAR |
S32G |
Arm® A53 (Neon) |
| S32R45 | Arm® A53 (Neon) | |
| Audio Identify emergency vehicle sirens in a noisy environment and localization of siren sound for ADAS |
S32G |
Arm® A53 (Neon) |
| SAF9100 | DSP AI/ML Accelerator |
S32 Family Hardware Accelerators
eIQ® Neutron Neural Processing Unit (NPU)
Dedicated AI/ML inference accelerator
- In-line dequantization, activation and pooling
- Built in tiny caching to reduce power consumption and reduce reliance on system memory speed
- Advanced multi-dimensional DMA for input and output formats, including striding, batching, interleaving, concatenating, etc.
- Weight decompression engine
- Configurable coupled memory
DSP
Signal processing, audio and sensor fusion
- Optimized for high-throughput vector math
- Used for deterministic DSP workloads
- Supports fixed-point and floating-point operations
Arm® Neon
Advanced single instruction multiple data (SIMD) architecture extension for Arm Cortex-A and Cortex-R series of cores.
Partner Ecosystem
eIQ Auto ML SDK FAQ
What machine learning model formats are supported by eIQ Auto?
eIQ Auto supports models from TensorFlow (Protobuf and Keras), PyTorch, as well as the ONNX interchange format and TensorFlow Lite (support varies depending on the underlying platform and backends).
What inference engines are supported by eIQ Auto?
eIQ Auto supports a range of inference engines including multiple open source offerings, hardware specific engines and a proprietary option developed under an ASPICE process. All the engines are unified under the common eIQ Auto model preparation workflow and runtime API.
Does eIQ Auto enable machine learning models to be executed on multiple processor cores?
Yes. The eIQ Auto runtime supports heterogenous execution with multiple runtimes and multiple processor cores. We can support many scenarios, configured using the eIQ Auto model preparation tools and executed with the eIQ Auto runtime libraries:
- A single model can be split, and the different parts can run either on the same core or different cores.
- Multiple models can run on different cores simultaneously.
Does eIQ Auto support execution on x86 processors?
Yes. The core eIQ Auto runtime libraries and certain inference engines are supported on x86 Linux, allowing you to fully prototype and evaluate your application on your host processor with eIQ Auto APIs before embedded deployment.
Does eIQ Auto support machine learning development beyond deep learning (neural networks)?
Yes. eIQ Auto has interfaces to support both deep learning and classical machine learning algorithms, as well as support for custom operations being executed by the eIQ Auto runtime.
Does eIQ Auto come with example ML models and applications?
Yes. The eIQ Auto installation includes a set of tutorials demonstrating the end-to-end development flow including model preparation on the host and verification of the runtime application, a set of demos showing more advanced features of the runtime, and a model zoo with additional application examples.
Design Resources
Learn about the eIQ Machine Learning Software and eIQ Auto and explore inference engine options for S32G, i.MX, i.MX RT and S32V processors.
Learn to implement and configure the NXP eIQ® Auto deep learning toolkit to optimize and implement DL without the need for customized hardware expertise.
Explore ways of dealing with these challenges and view an example of an optimized workflow for deploying deep learning in automotive production vehicles using the NXP eIQ® auto deep learning toolkit.
Additional Documents
eIQ Auto Deep Learning Toolkit Application Brief
Learn more about the processing efficiency, accelerated development and deployment workflows for AI automotive applications as well as how eIQ Auto Deep Learning toolkit assists your application development.
Read more