Image: © spainter_vfx/Stock.adobe.com
Start-ups from around the world will be showcased at the inaugural Intel Movidius Edge AI Incubator Open Day at Talent Garden Dublin.
Since it opened last October, things have been moving quite rapidly for Talent Garden Dublin, the latest facility in Europe’s largest network of digital innovation hubs. At the time, it announced that Intel would have a significant presence in the hub and today (17 April) the tech giant hosts one of its first major events at the Dublin City University campus.
The Intel Movidius Edge AI Incubator Open Day is the main event of one of Ireland’s first incubator programmes focused on artificial intelligence (AI) for computer vision. This morning, 10 promising start-ups from Ireland and farther afield will deliver a five-minute pitch followed by a question-and-answer session. There will also be an opportunity to meet the founders for in-depth discussions.
Below is some background on the companies presenting today, and what they’re creating with machine vision.
Aerodrums
Liverpool-based Aerodrums was founded in 2011 with the aim of creating ‘invisible’ drums for anyone to practise with. While it looks like the user is just playing ‘air drums’, Aerodrums’ advanced technology uses a high-speed camera and optical motion technology to pick up the motions of special drum sticks and sensors placed on the drummer’s feet. Silent to anyone else, the drummer can hear themselves playing on headphones.
Aerodrums also supports drumming in virtual reality using most mainstream devices. A highlight for the company came at CES 2018 when the kit got a moment in the spotlight during Intel’s pre-show keynote.
ApisProtect
Founded just two years ago, Cork-based ApisProtect has already generated quite a bit of ‘buzz’ internationally as one of the leading autonomous bee monitoring services around.
Started by Dr Fiona Edwards Murphy, Dr Pádraig Whelan and Andrew Flood, the company’s technology uses a combination of sensors to collect data on beehives including temperature, humidity, CO2, sound and movement. By gathering data from hives across the world, its machine-learning algorithms can tell beekeepers how healthy their hive is, rather than just producing raw data.
Dolotron
Dolotron, an engineering start-up from Croatia, specialises in software design, prototyping and quality assurance for the latest vision hardware.
Dolotron is working with the Intel Movidius Myriad platform to drive better product performance with lower power consumption, while enabling a smaller overall product footprint. The Neural Compute Engine also enables a dedicated hardware accelerator for deep neural network interfaces.
Evercam
Evercam’s timelapse and project management cameras service the construction industry, and its client list already features key players such as BAM, Kingspan, John Paul Construction and Savills.
Working out of the UK, Ireland and Pakistan, Evercam’s construction cameras offer features such as live site views accessible anywhere, marketing-worthy timelapse video and high-resolution images, and a simple, secure way to share the captured footage.
HiData
There are two major challenges in providing AI-based video analytics services. First, they require substantial computing power. Second, they need to be compliant with data protection regulation, such as the GDPR, which makes analysis of people captured in video footage via a remote server practically impossible.
To tackle both these hurdles, HiData is developing a box with embedded Intel Movidius machine vision technology that will enable users to perform video analytics locally at low power.
As part of the incubator programme, HiData set about recoding the technology it has developed so that it can be embedded on an Intel Movidius chip, and collecting data from trial partners. The next step is to annotate newly obtained data and perform further deep-learning neural network training to get this product market-ready.
Forestry Masters
Forestry Masters sets out to provide services to the global forestry business, identifying a particular opportunity in the bare-root forest nursery production sector. To provide some context, the US, Canada and Europe produce and manually process in excess of 3bn bare-root trees per year. Current production systems require 5,000 manual labourers and are reliant on the free movement of seasonal labour.
Forestry Masters is developing innovative products that will pivot on mechanical engineering, AI and vision systems in order to reduce labour requirements and increase productivity in this sector. The proposed grading machines and other equipment will assist nursery owners and provide support in the upstream and downstream phases of the production cycle.
Petronics (Sprite Robotics)
When three cat-loving electrical engineers met at the University of Illinois, they set upon a mission to create the best automated cat toy in the world. The result is Petronics by Sprite Robotics, which was founded in 2014 by David Jun, Michael Friedman and Dave Cohen.
Now a team of 10 (who own 11 cats), their flagship product is Mousr, a robot mouse that lets indoor cats enjoy the thrill of the hunt. Mousr is packed with sensors, actuators and intelligence and is always ready for kitty to initiate a game of cat and mouse.
Reflective Measurement Systems
Reflective Measurement Systems’ RetroTek devices evaluate the retroreflectivity and night-time visibility of road lines, markings and symbols. The company’s high-tech, cost-effective and user-friendly dynamic retroreflectometers address the needs of road transport and maintenance around the world.
While other devices are side-mounted, the RetroTek range mounts to the front of vehicles in order to be safer for the operator and other road users, as well as to allow the vehicle to ‘see’ markings across the full width of the traffic lane. The built-in reporting software provides interactive mapping and video for post-analysis and maintenance planning, and the company claims that using RetroTek offers 50pc savings in time, manpower and distance travelled.
Ambisense
Since 2014, Ambisense has been driving innovation within the environmental space. Ambisense’s Ambilytics technology platform optimises the delivery of environmental risk assessment on some of the world’s largest infrastructure projects across industrial, oil and gas, and waste verticals.
Ambilytics encompasses both internet of things and AI solutions, combining information from remotely deployed field devices with contextual data sources such as weather, satellite, geophysical and operational data to build machine learning models. These models identify the relationships, patterns, and drivers hidden within the data, enabling Ambisense customers to forecast and predict the behaviour of targeted environmental pollutants.
Tomra Sorting
Based in Citywest in Dublin, Tomra Sorting is an R&D hub of the wider Tomra group, headquartered in Norway. The Irish base’s focus is on designing the most technologically advanced optical sorting machines for the food industry. Its machines inspect and sort food produce in real time at capacities of up to 200 tonnes per hour.
Using the Intel Movidius Myriad technology, Tomra Sorting is looking to further develop ways of capturing images of foreign material that have been inspected by its machines, and to deliver these to the customer. This is a challenging problem as many optical food sorters operate in unforgiving environments with excess water, dirt and dust, which distort what the machine sees.
Updated, 1.56pm, 18 April 2019: This article has been updated to reflect that a company previously reported to be included in the Edge AI incubator, SparroWatch, was no longer participating at the time of the Open Day. SparroWatch’s entry has been replaced with Ambisense, which did showcase on the day.
