Hackathon gets set to take AI and computer vision to the edge

18 Oct 2018

An ‘AI at the edge’ processing card equipped with a Myriad 2 VPU pictured alongside the Neural Compute Stick. Image: Intel Corporation

A weekend hackathon at Talent Garden Dublin seeks to solve real-world problems using pioneering technology and AI at the edge.

Up to 100 artificial intelligence (AI) engineers are anticipated to descend on the newly launched Talent Garden Dublin hub at the Dublin City University (DCU) Alpha campus this weekend, 20 and 21 October. The reason: a hackathon to solve real-world problems using AI and computer vision.

“Computer vision itself can be looked upon as a specialism within AI, but this hack is looking at the next stage of AI development, ‘AI at the edge’, which is essentially AI models embedded within a device – no cloud, no Wi-Fi,” explained Neill Gernon, managing partner at Atrovate.

AI at the edge is a concept driven by leaders in the field such as David Moloney. Moloney’s company, Movidius (now part of Intel), is behind a USB stick designed to run AI models on-device – just one of the tools that will power this weekend’s projects.

As a form of AI, computer vision has been enabled by increasing computing power combined with the decreasing cost of high-quality camera equipment. However, video capture still equates to colossal amounts of data, and transferring that to the cloud for analysis is bandwidth-intensive. With AI at the edge, the analysis and insight can happen on-device, eliminating the unwieldy file transfer and also addressing other issues such as privacy and security.

‘We’ll be exploring what AI at the edge means for all types of computer vision applications and AI progress in general’

Atrovate, a Dublin-based protoypting and product lab specialising in AI, is just one of the stakeholders from Ireland’s AI community involved in this event, which brings together academic, corporate, start-up and investment interests. Gernon’s team will coordinate the hack and present interested hackers with computer vision challenges along with event partners from Intel and Insight, a Science Foundation Ireland research centre focused on data analytics.

“We’ll be exploring what AI at the edge means for all types of computer vision applications and AI progress in general, from AR and VR to IoT, medical diagnostics, autonomous vehicles and agriculture,” said Gernon.

Those who are new to software and data engineering and want to upskill in AI will get the chance to receive on-site training at hackathon workshops before they get down to building AI applications via accessible APIs.

“It’ll also give us a chance to highlight what’s next in AI and, of course, that means diverting attention to an area that is in massive need of more talent coming from third level but also the upskilling of existing engineering talent already in the market.”

Mixing it up

According to Prof Noel O’Connor, director of Insight at DCU, there’s plenty of appetite for this kind of work coming from third-level students.

“Our undergraduates in particular are realising that this is effectively a paradigm-changing technology irrespective of what field you’re working in,” he said, noting that about 30 undergraduate students from across different schools in DCU have signed up to participate, along with 10 postdoc researchers from his own group.

Also from O’Connor’s team, deep-learning expert Dr Kevin McGuinness and autonomous vehicles project leader Dr Suzanne Little will present talks during the event and be available as mentors.

What this means is industry AI engineers signing up for the event and academic AI researchers will be given the opportunity to share skills and ingenuity to build new and innovative tools. “Like any hackathon, you need a healthy blend of different disciplines: both the academic and industry, beginner and expert, business and technology – that’s what makes the magic happen,” said O’Connor.

‘You need a healthy blend of different disciplines – that’s what makes the magic happen’

There are few limits to what the hackathon teams might come up with, and they’ll be equipped with plenty of resources to enable them to dig in and get creative. Gernon explained that there will be a variety of accessible datasets, open source toolkits, hardware and APIs available for engineers, while event partners such as Intel have opened up access to key Movidius technologies such as the Neural Compute Stick and Myriad X vision processing unit.

Giving examples of what he expects to see worked on during the event, Gernon cited ideas such as computer vision applications for autonomous vehicles, computer vision to enable augmented-reality retail experiences in-store, and a medical device to analyse the breathing and movement of children with EB (known as ‘butterfly skin’) to monitor their health and detect changes or scratching.

“We’re also keen to let people explore, be creative, be innovative and be pleasantly surprised by what people come with,” said O’Connor. As he puts it, the event will give participants a fundamental understanding of AI at the edge, practical hands-on experience of working with this technology, and guidance on both its possibilities and its limitations.

“[We’ll] provide them with the tools and the knowledge to use those tools – where those tools are extremely powerful from a visual sensing perspective – and then [say]: go forth and see what you can come up with.”

The two-day AI hackathon takes place from 9am to 6.30pm on 20 and 21 October at Talent Garden Dublin, based at the DCU Alpha campus.

Elaine Burke is the host of For Tech’s Sake, a co-production from Silicon Republic and The HeadStuff Podcast Network. She was previously the editor of Silicon Republic.