Announcing New Integrations: Edge Impulse Optimizes NVIDIA AI for the Edge

Edge Impulse is pleased to announce that we have unlocked previously inaccessible NVIDIA AI capabilities for any edge device with NVIDIA TAO and Omniverse, alongside our launch of native support for NVIDIA Jetson Orin hardware. These developments further amplify what’s achievable at the edge for developers and enterprises who want to accelerate time to market and gain a competitive advantage with world-class AI solutions.  

With these new integrations, Edge Impulse provides the only way for developers to deploy NVIDIA technology directly to MCU and CPU. Engineers can speed up the use of large NVIDIA GPU trained models on low-cost MCUs and MPUs with AI accelerators, while accessing a powerful set of tools to create digital twins, synthetic datasets, and virtual model testing environments.

The NVIDIA TAO Toolkit: A game-changer for enterprises

NVIDIA TAO delivers a low-code, open-source AI framework to accelerate vision AI model development suitable for all skill levels — from beginners to expert data scientists.

With the NVIDIA TAO Toolkit integration, Edge Impulse’s enterprise customers can now use the power and efficiency of transfer learning to achieve state-of-the-art accuracy and production-class throughput in record time with adaptation and optimization. This is integrated directly into the Edge Impulse platform for any existing object detection project, available for all enterprise users today.

The TAO Toolkit provides a faster, easier way to create highly accurate, customized, and enterprise-ready AI models to power your vision AI applications. The open-source TAO toolkit for AI training and optimization delivers everything you need, putting the power of the world’s best Vision Transformers (ViTs) in the hands of every developer and service provider. 

Built on TensorFlow and PyTorch, the NVIDIA TAO toolkit uses the power of transfer learning while simultaneously simplifying the model training process and optimizing the model for inference throughput on the target platform. The result is an ultra-streamlined workflow. Take your own models or pre-trained models, adapt them to your own real or synthetic data, then optimize for inference throughput. All without needing AI expertise or large training datasets.

New NVIDIA TAO model options now available in Edge Impulse

The integration of NVIDIA TAO into Edge Impulse means that engineers can finally utilize NVIDIA’s industry-leading AI models on hardware outside of that offered by NVIDIA, a capability exclusively provided by Edge Impulse.

Over 100 accurate, custom, production-ready computer vision models are accessible via the Edge Impulse and NVIDIA TAO Toolkit, allowing engineers to seamlessly deploy to edge-optimized hardware, including the Arm® Cortex®-M based NXP I.MXRT1170, Alif E3, STMicro STM32H747AI, and Renesas EK-RA8D1. 

"The advent of generative AI and the growth of IoT deployments means the industry must evolve to run AI models at the edge,” said Paul Williamson, senior vice president and general manager, IoT Line of Business, Arm. “NVIDIA and Edge Impulse have now made it possible to deploy state-of-the-art computer vision models on a broad range of technology based on Arm Cortex-M and Cortex-A CPUs and Arm Ethos™-U NPUs, unlocking a multitude of new AI use cases at the edge." 

Access the NVIDIA TAO documentation to learn more.

Users can configure multiple parameters within Edge Impulse-generated TAO models.

NVIDIA Omniverse: Get to market faster with synthetic data generation 

Edge Impulse is also launching our NVIDIA Omniverse integration, allowing for synthetic data and testing environments to enable faster time-to-market in key business verticals.

With AI and machine learning, robust, quality data is crucial for building accurate and effective models. In some scenarios, however, obtaining sufficient real-world data can be costly, time-consuming, create privacy concerns, or bring other challenges. Synthetic data is proving to be a critical breakthrough for industries operating in complex industrial, remote, or sensitive environments.

With the NVIDIA Omniverse integration, users can access the develop custom synthetic data generation pipelines to rapidly generate highly realistic, physically based datasets tailored to train computer vision models for viable real-world models. 

In addition, models created on the Edge Impulse platform, with real and synthetic data, can now be validated with NVIDIA Omniverse digital twin functionality. You can now rapidly iterate and gain insights into what works best for your application. The digital twin allows you to simulate sensor and model behavior, test MCU compatibility, and more. This capability helps teams optimize and fine-tune AI models before deploying them, saving time and resources in the development process.

The NVIDIA Omniverse integration lets users utilize trained models directly within the Omniverse GUI; you can see real-time object detection bounding boxes within the synthetic environment and validate your model’s accuracy. It can also directly pull trained Edge Impulse models into local computers or edge devices, and get inferencing results in real-time from the 3D Omniverse environment. The integration simplifies data uploads with its upload portal — just specify an API key and directory and you can upload the Replicator data into your Edge Impulse project.

The NVIDIA Omniverse and Edge Impulse integration is available for free to all developers. For those who wish to utilize the Cloud options, you can do so by hosting in your own AWS server instance, as well as with the applicable NVIDIA GPU hardware.

Also see Omniverse Replicator in action with this Edge Impulse project for detecting surgical tools left behind post-procedure.

Synthetic image from Omniverse Replicator

Seamlessly deploy to the edge with NVIDIA Jetson Orin — and more!

After training and validating your TAO model, with or without synthetic data, you can now deploy it to any device. This makes the model run without an internet connection, minimizes latency, and runs with minimal power consumption.

Edge Impulse supports six deployment options, including powerful NVIDIA Jetson Orin-based hardware. NVIDIA computer vision models deploy faster on Jetson Orin. You can quickly solve industrial business problems with the hardware you know, and the ML development tools you need.

Getting started with Edge Impulse's NVIDIA integrations

All the tools you need to successfully deploy your AI model are available and ready for  use. Setting up Edge Impulse with NVIDIA TAO and Omniverse is easy. For NVIDIA Omniverse, get started here or access NVIDIA TAO docs

With these innovations at your fingertips, we can’t wait to see what you create. If you’re working on a professional project, we invite you to book a demo with our team or sign up for a free trial of our Enterprise platform


Are you interested in bringing machine learning intelligence to your devices? We're happy to help.

Subscribe to our newsletter