To further help innovators and developers attending InnoVEX 2019, one of the IoT forum sessions was specially held by Arm. Featuring keynote speakers who are active in the community, both from Arm and users of Arm's services; the session discussed how they can achieve their goals.

Leverage Ecosystem to Develop Products

Starting the session is Mr. Alessandro Grande, Ecosystem Manager of Arm. In his speech, he mentioned that with the IoT explosion in recent years, the number of devices containing Arm chips has also grown exponentially. This is also a great opportunity for both startups and companies to enter the IoT industry; especially as companies such as Arm are also providing their own development tools for innovators to use and then join their community.

Mr. Grande quoted a report by Gartner that predicted there will be more than 500 smart devices in every home by 2022; further enhancing the scope and opportunities in ioT. If autonomous cars are also put into consideration; this means there will be trillions of devices which are collecting data. The existence of all of these different types of devices will create a tremendous amount of data that will necessitate autonomous decision making on the hardware side. Regardless of what they are building, developers and innovators both need to think about how to process the collected data. His suggestion is using edge processing.

More people, startups, and companies have actively participated or are considering migrating to machine learning in edge devices. This trend is further enhanced by the ubiquity of smart phones which uses CPU or CPU & GPU to do machine learning on the device. One example for this is how many smartphones of today can use facial recognition to unlock the devices in question. The mentioned example also shows how certain vital functions should not be left to be processed in the cloud as well as the necessity for edge or device based computing. Machine learning on the edge is also a great opportunity as the number of data that needs to be processed continues to grow. This means there's an opportunity for everyone building devices to actually process the data without sending it to and from the cloud.

Another, more key factor in promoting edge computing is the issue of bandwidth. With trillions of data all transmitting and data, current communication infrastructure will not be able to handle the load. Continuous data transmission also means that there are additional vulnerabilities that can be exploited by malicious agents. Keeping data on the device will greatly reduce exposure to attacks.

Strength through Community

As stated, one of Arm's main strengths is in their community which includes developers and innovators. One of them is Mr. Seth Welday, Product Marketing Manager of SEEED Studio; a member of the Arm ecosystem. As an IoT hardware enabler, SEEED Studio aims to get IoT hardware into the hands of developers.

As the types of IoT devices continue to grow along with the functions they perform; customizability will be a major point. Mr. Welday said that developers can benefit from Arm more specifically because they do not manufacture their own chips. In addition, it becomes easier to create a feedback loop within the ecosystem as any new technology that is embedded in the chip can be modified following feedback from the users.

Often times, startups that want to get into hardware are limited by manpower and especially for hardware startups, by the technology limitations. Joining the community will help bridge technology gaps for hardware prototyping; especially as some development boards are specifically built to work with certain ecosystems.

Mr. Welday stated that perhaps one of the most useful, but less appreciated parts of the community are the tutorials and documentations. While often not considered to be an ecosystem on their own, people regularly contribute online and the contents offer great insights to kickstart projects. In addition, official documentation from the companies and forums such as Stack Overflow are also great sources for troubleshooting and preventing human errors.

There are many IC companies and they have specific tools to help developers program and debug for development with competitive prices. They all have solutions and can help startups get to market more quickly because they have tools built around them. When choosing partners for such a large and growing industry, it is vital to choose one that fits the requirements. When companies have all these tools to give to their developers, it allows the developers to quickly prototype, get to market, scale, and succeed. After that, ideally and hopefully they will return for future projects. If not, they will still be part of the ecosystem and might return in the future.

Computing on the Edge

One area where edge computing is preferable to cloud computing is in visual AI. Whether for face recognition or image recognition, rapid detection and response are both necessary for the system to be considered a success. Another member of the Arm ecosystem is DT42, an end-to-end solution provider for AI with a focus on visual AI such as smart cameras. The speaker for this session is Mr. Kevin Su, Project Manager of DT42.

Mr. Su stated that DT42's vision for the future correlates to the perspective of Arm in that more computing should be done on the device rather than the cloud or servers. Focusing more on cloud computing will allow for split-second decision making; supported by inference & model optimization.

As the model is trained, it becomes more efficient and will be able to do more and the importance of versatile hardware and reliable partner become more prevalent. Multiple use cases as was aimed by DT42 means users can customize and tailor the sensors to do whatever is needed depending on how the inference model works.

Currently, the 2 main use cases for smart cameras are classification and detection. With classification, the video is cut into frames and the content is identified; something commonly done by smart cameras of today. In contrast, detection requires a combination of classification and localization; tracking a person and classifying them over several frames and ascertaining their destination. This will create new possibilities such as people counting, fall detection, and more; opening an even broader spectrum of use cases.

A reliable partner such as Arm which has various offerings is greatly beneficial to solution providers as it can give them confidence that they can move within different disciplines easily.

To watch the full forum session, visit our YouTube channel here.