Understand how embedded AI is changing the way we shop


Case Study: Embedded Artificial Intelligence Unleashes Smart Shopping

An effective IoT business model delivers added value to customers and allows seamless product monetization. 

8 IoT Business Models for Connected Product Strategy

<h2 align="center" id="isPasted">Leading companies present intralogistics technology trends</h2><h3 align="center" id="isPasted">Future of Intralogistics Days 2022</h3>Concentrated logistics power without the waffle – at the virtual&nbsp;<a href="https://www.future-of-intralogistics.com" rel="noopener noreferrer" target="_blank">Future of Intralogistics Days 2022</a> conference on March 9th and 10th, 18 famous companies will be presenting the latest intralogistics technologies and best practices. The free event is aimed at experienced intralogistics experts in the European B2B sector. The focus of the 20-minute sessions will be new and future trends and examples of applications for intralogistics processes in storage environments, e-commerce and production.<img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNjQ1NzA3NTk1OTUzLWFsbGVfbG9nb3MxLmpwZyIsImVkaXRzIjp7InJlc2l6ZSI6eyJ3aWR0aCI6OTUwLCJmaXQiOiJjb3ZlciJ9fX0=" class="fr-fic fr-dib"> Driverless transport systems, sensor technology and digital infrastructure – the first day of the event is defined by mobile robotics. Exciting application examples, technology presentations and live discussions will illustrate the results that can be achieved with mobile robots and the demands that implementation will bring. Everything is user-related and practical. The same is true of the second day. Here, it’s all about wearables, software and materials handling. There will be speakers from<ul><li>Siemens</li><li>SAFELOG</li><li>SICK&nbsp;</li><li>FlexQube&nbsp;</li><li>NIMMSTA&nbsp;</li><li>Picavi&nbsp;</li><li>cellumation&nbsp;</li><li>Phoenix Contact&nbsp;</li><li>REMIRA&nbsp;</li><li>Sherpa Mobile Robots&nbsp;</li><li>MHP - A Porsche Company and <a href="https://future-of-intralogistics.com" rel="noopener noreferrer" target="_blank">more...</a></li></ul>Interested logistics experts will find the full program at <a href="https://future-of-intralogistics.com" rel="noopener noreferrer" target="_blank">www.future-of-intralogistics.com</a>. The event will be in English.

Concentrated logistics power without the waffle – at the virtual Future of Intralogistics Days 2022 conference on March 9th and 10th, 18 famous companies will be presenting the latest intralogistics technologies. The free event is aimed at experienced intralogistics experts in the B2B sector. 

Future of Intralogistics - Days 2022

Deep neural networks have always been at the forefront of the implementation of computer vision applications. However, the increase in ML computation from cloud to edge introduces new challenges in a resource-constrained environment. For DNN, there are multiple layers between the input and output as the information from the input is represented through nonlinear functions. The existing challenges of increasing energy consumption due to continuous computation of the incoming data and reduction in accuracy with latency constraints have led the researchers from the University of California at Irvine and Northeastern University to develop a robust framework to overcome these issues. &nbsp;Before jumping to the new BottleFit framework’s proposition, let us look at the existing design flow approach. The recent developments for solving the computation requirements focused on reducing model complexity and designing lightweight models like MobileNet and MnasNet. But these innovations came at the cost of significant accuracy loss of up to 6.4% compared to the large CNN models ResNet-152.&nbsp;In this paper, the innovators have introduced a holistic approach and alternative solution to edge computing wherein, through split computing, a larger DNN is divided into head and tail models to be executed at the mobile device and edge server, respectively. The BottleFit framework not only aims to not only reduce power consumption and latency but achieves high accuracy even with strong compression rates. The critical aspect while designing the framework is to maintain the tradeoff between head model compression and tail model accuracy. Usually, with high compression rates, there comes a severe decrease in accuracy; however, this novel architecture and custom design flow of improved strategies goals towards maintaining high accuracy and bottleneck compression.<h2>Effective and Efficient Split Computing in DNN</h2><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1643392197564-Split+Computing+System.png" style="width: 800px;" class="fr-fic fr-dib" alt="Split Computing System in DNN">Split computing: (Left) The original pre-trained model is introduced with a bottleneck and then split into head and tail sections. (Right) Split computing system used in this experimentation setup. [Image Credit: Research Paper]<div is-sapling-overlay="true" style='pointer-events: none; overflow: hidden; padding-bottom: 2px; z-index: auto; position: absolute; left: 0px; top: 397.1px; height: 63px; width: 766px; font-style: normal; font-variant: normal; font-weight: 400; font-stretch: 100%; font-size: 14px; font-family: "PT Serif", serif; text-align: center; text-transform: none; letter-spacing: normal; word-spacing: 0px; tab-size: 8;'> </div>Understanding the split computing model used in this experimentation setup: the mobile edge device captures a high-resolution image given to the head model that performs compression and sends it to the edge server through wireless channels. The results are received at the edge server, which is fed to the tail model to provide inference output and transmitted wirelessly to the mobile device.&nbsp;In split computing, the training process is not split but done offline. A bottleneck is defined to split the original model into head and tail sections. The bottleneck is introduced through encoder and decoder structures within the original model in the new proposed methodology. The layers are operated in the encoder-decoder approach of mapping the model input to the intermediate output and training to execute the downstream task without significant loss in accuracy. Through this approach, the researchers could influence the tradeoff between computing load on the mobile device and complexity and compression gains. All these factors substantially affect the overall performance of energy consumption, low latency and improved accuracy.“Thus, when introducing bottlenecks, we need to carefully design the encoder and decoder; choose the bottleneck placement in the head model, and preserve the accuracy of the unaltered original model as much as possible,” the researchers explain.&nbsp;<h2 dir="ltr" id="isPasted">Performance of BottleFit Framework– Outperforming Existing Design Flow Approaches</h2><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1643392214341-Experimental+Setup.png" style="width: 800px;" class="fr-fic fr-dib" alt="Experimentation Setup of BottleFit">The hardware setup showing the Raspberry Pi 4 General Purpose Single Board Computer and NVIDIA Jetson Nano (GPU-based) [Image Credit: Research Paper]<div is-sapling-overlay="true" style='pointer-events: none; overflow: hidden; padding-bottom: 2px; z-index: auto; position: absolute; left: 0px; top: 436.662px; height: 63px; width: 766px; font-style: normal; font-variant: normal; font-weight: 400; font-stretch: 100%; font-size: 14px; font-family: "PT Serif", serif; text-align: center; text-transform: none; letter-spacing: normal; word-spacing: 0px; tab-size: 8;'> </div>The experimentation of the framework is carried out on one of the most widely used TinyML applications, image classification, showing the BottleFit framework achieving over 70% data compression while only losing 0.6% on the accuracy front on the ImageNet dataset. The existing state-of-the-art design flow approaches, such as SPINN lose up to 6% in the accuracy for high compression rates. These evaluations were done on two famous embedded electronic devices, one popular for edge AI applications and the other known for its general-purpose tasks, NVIDIA Jetson Nano (GPU-based) and Raspberry Pi (GPU-less), respectively.&nbsp;When we discussed improving the energy efficiency and lower latency, the BottleFit framework deployed on NVIDIA Jetson Nano shows the reduction of both parameters by up to 54% and 44%, respectively. While the exact model running on Raspberry Pi demonstrated a decrease of 40% and 62% on the power consumption and overall execution time. Along with this, it is noticed that the size of the head model executed on the mobile devices is 83 times smaller.<h2 id="isPasted">Summary of BottleFit Framework</h2>Summarizing the new BottleFit framework that improves power consumption at high compression rates and maintains accuracy loss:<ol><li dir="ltr">A novel framework of multistage training strategy that archives high accuracy with strong compression rates through properly designing the encoder and decoder structures to introduce bottlenecks in the original pre-trained model.</li><li dir="ltr">The evaluation of Jetson Nano board commuting with the edge server using Wi-Fi and LTE while RPi using LoRa shows considerable improvements in the overall performance.&nbsp;</li></ol>For more details on the research methodology, check out the <a href="https://arxiv.org/abs/2201.02693" target="_blank">open-sourced research paper</a>.&nbsp;<h2 dir="ltr">Reference</h2>Yoshitomo Matsubara, Davide Callegaro, Sameer Singh, Marco Levorato, Francesco Restuccia: BottleFit: Learning Compressed Representations in Deep Neural Networks for Effective and Efficient Split Computing. DOI <a href="https://arxiv.org/abs/2201.02693" target="_blank">arXiv:2201.02693</a> [cs.LG]M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 4510–4520.M. Tan, B. Chen, R. Pang, V. Vasudevan, M. Sandler, A. Howard, and Q. V. Le, “MnasNet: Platform-Aware Neural Architecture Search for Mobile,” in Proceedings of the IEEE Conf. on Computer Vision and Pattern Recognition, 2019, pp. 2820–2828.

BottleFit Image Classification through Advanced Split Computing Strategy with Jetson Nano and Raspberry Pi 4

BottleFit framework applied DNN models in image classification achieves 77.1% data compression with up to 0.6% accuracy loss on the ImageNet dataset.

Advanced Split Computing Strategy to Achieve High Accuracy under Strong Compression Rates

As a novel coronavirus swept the world in early 2020, thousands of software developers began working from home. Many did so on short notice, under difficult and stressful conditions, feeling frustrated and insecure. However, others used this remote work opportunity with optimism and turned their temporary limitations into new opportunities.&nbsp;A <a href="https://www.rinf.tech/" target="_blank">rinf.tech</a>&nbsp;senior software engineer – <a href="https://www.linkedin.com/in/emanueloprea/" rel="noopener noreferrer" target="_blank">Emanuel Oprea</a> – is one of those who used the pandemic and self-isolation to indulge in a pet project that subsequently became a full-fledged solution enabling users to use VPN client software on many different devices at the same time.We’ve talked to Emanuel about the drivers of his solution development and how his product evolved from a simple script into a market-ready router VPN management solution.<h3 dir="ltr">• &nbsp; &nbsp;Emanuel, can you please share a story behind your technology solution?</h3>Once upon a time, my wife saw an enticing trailer of a new show and wanted to watch it through a streaming service. In particular, she wanted to watch the show on a large TV screen in the living room, lolling in the comfort of the couch. However, she couldn’t do it due to the streaming service geo limitations in Romania.An active VPN connection is needed to bypass those limitations and watch certain shows that are unavailable in your country. &nbsp;While this issue can be easily solved with a VPN on a laptop or a smartphone, watching the same show on TV is an entirely different story.That’s exactly how I stumbled upon a user's need for streaming content blocked by ISPs on a smart TV or other devices like gaming consoles, etc. And, since I had more free time due to the pandemic and remote work, I decided to help my wife find a solution to enable her to watch blocked shows through VPN on our smart TV.<h3 dir="ltr">• &nbsp; &nbsp;And what was your first step?</h3>Like anyone in my situation, I searched on Google and tried to find an existing product. And then, it was already mid-2020, and I’ve ended up spending quite a lot of time looking for software already out there on the market. Almost all the solutions I found only covered PC or mobile/tablets. The ones that directly tackled my issue proved to have more limitations than benefits.One of the initial options I discovered was a VPN provider that also has its own custom firmware. Ideally, one could install it on their router, but after further investigation, I found that it only supports five router models, and of course, due to Murphy’s law, mine wasn’t supported.The other option meant buying a new router from a company specialized in router firmware customization. After device acquisition, the company would install a proprietary VPN client on my future router. The downside of this was the final cost of the router (versus retail price) and the dramatic increase in shipping time. It also proves to be a warranty nightmare since they would have to open the package to operate changes on the device before finally sending it to me.Since I already had a router, I didn’t want to spend more money buying a new one, and I also wanted to keep my existing VPN subscription. So, I said to myself:<blockquote>“C’mon, Emanuel, you’re a software engineer! Why don’t you try to build an ultimate router VPN solution and help others easily overcome streaming platforms’ limitations with an added security boost to the internet connection?”&nbsp;</blockquote>That’s how I took on this new challenge which eventually resulted in ShieldMe solutions.<h3 dir="ltr">• &nbsp; &nbsp;And what did you do next?</h3>I started with writing a simple bash script to try to connect my router to VPN. It was a proof-of-concept that worked well. However, it could only be used by my peers, tech-savvy people. A solution like this would prove to be a bottleneck for mass adoption, much like the already existing ones. So, I understood I had to make something more user-friendly and polish the solution further. This translated into my next goal: building a simple and intuitive app that would be available for “virtually” any router.The app started small, with just a couple of simple screens with basic options. Then, thankfully, my wife gladly accepted to act as my trusted quality assurance department. With her precious, albeit intense (but very constructive) feedback, the initial MVP was ready after a few iterations, and I launched my VPN router client in April 2021. However, the solution had its limitations as an MVP, so that I couldn’t call it “universal” just yet.&nbsp;<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1637669578420-Showcase-ConnectionHistory.png" style="width: 720px;" class="fr-fic fr-dib">Initially, users had to know how to run an installation script to use my VPN client, discouraging most people from trying it out. As such, to fine-tune the initial solution, I started working on transforming it into a set of multiple user-friendly tools: a router manager and a VPN manager. I really wanted users with limited to no tech background to be able to use it with no pain in the neck.That’s what pushed me to build and release a complementary router automation app. With it, users can now install custom firmware on their router and set it up automatically through a user-friendly interface: just two actions, and they’re good to go.<h3 dir="ltr">• &nbsp; &nbsp;Can you tell us more about how a non-tech user can benefit from your router VPN solution?</h3>Well, to name just a few: users can access media content anywhere and anytime, privately and securely, through an app that is easy to install and use. They can enjoy streaming with no geo-restrictions or interruption, thanks to the fast connection with no speed or bandwidth limits on their devices. They can also use the internet much safer, thanks to the underlying connection encryption and anonymization. Usually, VPN services limit the number of connections to their servers. Still, by using my client directly on the router, you essentially allow all devices behind it to count as a single connection. And more exciting developments are on the way, as I am currently working on a mobile app for my solution!<blockquote>But I believe that the main benefit is that my solution helps democratize access to streaming services and content around the world.</blockquote><h3>• &nbsp; &nbsp;What technologies and frameworks did you use to build your product?</h3>I used JavaScript for the app front-end, PHP for router back-end, and .NET for interfacing with VPN APIs.&nbsp;Feel free to learn more about Emanuel's solution on the <a href="https://shield-me.net" rel="noopener noreferrer" target="_blank">ShieldMe website</a>.

Emanuel Oprea, a software engineer at rinf.tech, shares a story of how he accidentally identified a user near during the last year's Covid-19 lockdown and built a market-ready router VPN manager to help democratize access to streaming content and services.

From a simple bash script into a full-fledged router VPN management solution

This article provides a comparison between NB-IoT and LTE-M protocols and explores how NB-IoT facilitates energy-efficient and latency-tolerant IoT solutions for hard-to-reach locations.


embedded software

Latest Posts

Case Study: Embedded Artificial Intelligence Unleashes Smart Shopping

8 IoT Business Models for Connected Product Strategy

Future of Intralogistics - Days 2022

Advanced Split Computing Strategy to Achieve High Accuracy under Strong Compression Rates

From a simple bash script into a full-fledged router VPN management solution

How NB-IoT facilitates energy-efficient and latency-tolerant IoT solutions

Profiles