Description

Imagine a new world where you could add intelligence to all your embedded devices simply by collecting the right data to fully characterize your problem, that world is here.

Join this keynote to to celebrate with us all the amazing ecosystem achievements of this last year, and hear directly from Zach and Dan some of the biggest advancements characterizing the embedded ML space today.

Description

Join this keynote to hear from Steve Roddy, what to expect from the next generation of Arm-based hardware devices.

Description

Oura, is fully dedicated to improving the way people live their lives. They are achieving this by developing advanced technologies with the Oura ring to monitor your body’s pulse, movement, temperature etc and deliver a holistic picture of your health via the Oura app. ML-powered algorithms are resulting to be the key to their missions.

In this presentation, Xi will talk about how Oura leverages machine learning infrastructures and technologies to innovate in the health space by improving the accuracy and quality of the health information they provide to their users.

Description

Join this conversation to hear Ray's vision on the future of smart and connected edge devices.

Ray's Bio:

Ray Ozzie is a software entrepreneur with a particular passion for communications software and infrastructure.  Having created numerous products over the years including Microsoft Azure and Lotus Notes, his early career was shaped by his work on Lotus 1-2-3, Symphony, and VisiCalc.  He was CTO and chief software architect of Microsoft, is currently CEO of Blues Wireless, and also serves as a director of Hewlett Packard Enterprise, Balena, and Safecast - a global citizen science initiative founded in the wake of the Fukushima disaster.

Description

Machine learning has traditionally been seen as something that happens in data centers, requiring a lot of resources, so that only large organizations benefit from the technology. Running models locally on devices gives us a chance to share the benefits more widely, but also to reimagine how we design algorithms, hardware and software. In this talk, we discuss how the edge is fundamentally changing how we apply machine learning and explore some of the open challenges and opportunities to transform user experience at the edge.

Description

In this session Zach Shelby will introduce Luka, CEO and founder of Irnas, and together they will explore how by focusing on the value of data, Irnas has become one of the most innovative design houses.  The transition from a generalist IoT design house to a “Data-First” approach has brought a lot of successes.

Irnas today together with Edge Impulse, is able to help customers add value to their products by making devices smarter and able to make sense of the data around them.

Description

Join this keynote to learn about Zach's vision on the future smart IoT devices.

Description

Join this conversation with Pete Bernard to explore how 5G is accelerating embedded ML adoption at scale, while unlocking more revolutionary use cases across the industry.

Description

Technologists have a huge role to play in driving positive change in our planet’s ecosystems. Scientific developers need our partnership to solve problems to preserve our biodiversity and drive sustainability.

Join Sarah Maston, founder of Project 15 from Microsoft as she discusses how cutting edge technology in IoT and ML are making impact to accelerate innovative solutions for Earth.

Description

Join this industry panel to gain deeper insight into the impact embedded intelligence is having on today's enterprises. Panelists will share their views on the current landscape and the future of edge AI as well as unpack ways organizations can take it into production now.

Description

Join this session to hear some of the highlights of the opening day and explore the agenda for the second day of Imagine.

Description

Join this panel hosted by Sally Ward-Foxton from the EE Times, to hear from industry experts what the needs and challenges of modern enterprises are, and what we should all be doing to be ready for a smarter future.

Description

Join this workshops with TI to explore how embedded ML can enable enterprises to prepare for the future.

Description

Detecting objects is one of the most challenging problems in computer vision. In this session, we will show a complete example and discuss some of the strategies we have developed on the basis of Himax WE-I Plus platform.

The workshop starts off with model selection, followed by data preparation, training procedures and finally deployment with TFLite. Participants will learn how to deploy the tinyML object detection model  on their devices.

Description

The rise of intelligence on the edge is inexorable. Nvidia’s Jetson family of modules provides a GPU-accelerated AI platform for edge computing. With Edge Impulse and Nvidia application SDKs, developers are armed with the tools to create embedded ML applications for Jetson(s) frictionlessly.

Ready for production? Nvidia’s Triton inference server helps you deploy your AI solutions at scale and orchestrate multi-tenant ML model execution with optimized CPU/GPU configurations on the edges. From home robots to machine vision, from AgTech to nanosatellite, Jetson-powered AI use cases are penetrating all the industry verticals.

Description

Introducing new crop varieties or introducing new cultivation methods requires intensive trials to irrigate and monitor the performance of each plant. To add to the further challenges trials sites are often based in isolated location for security and privacy.

MARV-Bot is an agricultural trials robot designed to support remote operation at trials sites. As such MARV is capable of irrigation of the crops and the monitoring and classification of individual plants.This communication can be provided back to the control center using either 5G or SatCom networks depending on location. The remote operation requires a low power efficient ML applications to be able to identify each crop type and geotag each of the plants for cataloguing and growth monitoring.

This work shop will outline how Edge based ML can be deployed for agricultural robots. We will explore the challenges of agricultural robotics and how ML can help address these challenges in general, and more specifically why edge ML can help achieve a more efficient and longer lasting implementation (lower power, lower bandwidth needed per link etc)

Finally we will wrap up with the MARV Bot prototype recreation which as implemented on the Sony Spresense system.

Description

This hands-on workshop will show how to use the Edge Impulse Linux CLI to gather and annotate data and then train and deploy a hard hat detection model to the MaaXBoard Mini. This dev board is based on NXP’s 8M Mini, an excellent architecture for running object detection inference. We’ll also harness the GPIO on MaaXBoard’s raspberry pi compatible 40-pin header to set off an alarm if the model detects someone isn’t wearing their hard hat. Safety first!

Description

Join this workshop to learn how to develop machine learning applications on both short range nRF5340 and long-range nRF9160 Nordic radio devices. We’ll show you how easy it is to get started with ML development with out of the box applications from the nRF Connect SDK using data captured from various sensors. The workshop will be split into two 30-minute slots where the first part will focus on short range with support for Bluetooth LE and the second part on cellular (LTE-M/NB-IoT) ML offering.

Description

Syntiant designed an artificial intelligence neural decision processor from scratch. Our purpose-built neural decision processor (NDP) provides 10-100x throughput while consuming 10-100x less power for the same task, relative to MCUs. This can be achieved thanks to having the neural layer operations baked into the architecture itself. The user-friendly architecture does not even require a compiler which will be difficult to maintain and support.
‍
Developers can focus on developing and managing machine learning models using the Edge Impulse front end, and simply download the model package and deploy it on the board. In collaboration with Edge Impulse, Syntiant is delivering two orders of magnitude improvement in throughput and power efficiency with an easy-to-use platform that enables developers to deploy strong machine learning algorithms anywhere.

Description

Discover how easy it is to integrate Edge Impulse solutions in your Microchip project

Description

Join Massimo Banzi, co-founder of Arduino, to hear the latest about Arduino and how it is enabling more developers to deploy ML applications on connected devices.

Description

This panel will highlight some of the efforts helping to drive innovation in embedded ML. The discussion will focus on the work of the developer community, emerging trends and applications, as well as the adoption of tinyML-compatible hardware and tools.

Description

This panel will dive into the current state of embedded machine learning, identifying disruptive developments, addressing obstacles and opportunities, and exploring what's next for data-driving engineering.

Description

Embedded machine learning becomes much easier and accessible to beginners when using Codecraft graphical programming and Wio Terminal.Simply drag-and-drop blocks, acquire data, train and deploy models, Codecraft will bring you a more vivid machine learning experience than ever.Whether you are only at the stage of transforming hardware into a smarter device, but don’t know how to start to execute, or you came up with the idea, but lack the developing skills with a practical way to implement and accelerate your next IoT projects, you will find this Graphical Programming platform to be one of the best learning and training resource for of your TinyML journey.Powered by Edge Impulse, TinyML community, and Graphical Programming platform, we are aiming at reducing the computational space and power required by machine learning models, Edge AI is fitting artificial intelligence applications into smaller, less powerful computers, even ultra-low-power microcontrollers, and less required developing skills.

Description

This workshop will teach attendees how to build their own AI-powered bird watcher from a Raspberry Pi and a camera!  As birds pass in front of the camera, an Edge Impulse AI model will try to determine what species it is, and send a picture of the bird (and it’s predicted type) to the user via a Telegram bot.  And by using balenaCloud to deploy the software onto the Pi, the process is easy and repeatable, allowing developers to continually update devices and add more units.  This is a great project for schools and citizen science, where added data can improve the model and the whole fleet of bird cams can get smarter.
Attendees will need a Raspberry Pi 3 or 4, a Pi Camera or USB Webcam, and an SD Card and Power Supply in order to participate in the hands-on build, or, you can simply follow along as we talk through the lab and can replicate it all later on your own.

Description

Join Alex during this session for a chance to learn about the latest embedded ML projects being built by the community.

Description

Join this workshop, for a chance to learn step-by-step what it takes to to build an artificial nose using off-the-shelf, affordable and open hardware. During the second half of the workshop, you will also learn how to add IoT capabilities to the device in order to support both disconnected and connected scenarios.

Description

MLOps is a set of tools and techniques. to train, deploy and manage in large teams and powerful servers. However, parallel to the development of these large models, there has been a growing field of TinyML which tries to reduce model sizes and complexities so that they can be deployed on small, low power and low memory microcontroller devices.

TinyML brings its own set of challenges to MLOps when it comes to building scalable pipelines to train, optimize, deploy and maintain these models. For instance these models need to be small and have low inference latency but at the same they need to be accurate. In this workshop we will talk about what Tiny MLOps is and the challenges that Tiny MLOps systems have. We will also give an overview of the Tools available for Tiny MLOps and demonstrate how to build a Tiny MLOps pipeline for deploying a classification model using Edge Impulse.

Description

Deploying an IoT solution on the edge comes with inherent risks. Is the firmware bulletproof? Is the enclosure secure? Is the power source reliable? Then you add ML to the mix and you have to hope your model is "good enough" to be objectively successful.

However, what if your remote deployment could continue to learn, feed new data back to Edge Impulse, and in turn continually improve the model? What if you could then programmatically update the model file remotely over cellular, without manual intervention?

Using the Notecard and Notehub.io from Blues Wireless, this is now a reality. Join us in this workshop as we demonstrate the basics of the Notecard on a Raspberry Pi 4 model B, along with an end-to-end solution of dynamic model updates in the field.

**To follow along the workshop you will need the Blues Wireless Raspberry Pi Starter Kit

Description

Data augmentation is an incredibly useful technique in machine learning that can multiply your dataset to train a more robust model. We will start with a limited, pre-existing dataset (images of electronic components) and use it to train an image classifier. We will examine flaws in the model by looking at saliency and class activate maps from validation and test images. From there, we will use Google Colab to perform various transforms on the original images to create a more robust dataset, and use it to train a new model. Finally, we will discuss how data augmentation can be applied to non-image data.