Eugenio Culurciello, co-Founder of TeraDeep, wants to give vision to machines. TeraDeep, described by Gigom as “straddling the intersection of two very big trends — deep learning and the internet of things”, provides high-performance computing solutions for real-time complex-data analytics. Their solutions are a clever combination of machine-learning algorithms, cloud-intelligence software, custom system-on-a-chip, special embedded computers and server-side devices.
We caught up with Eugenio ahead of his talk at the Deep Learning Summit next week.
What led you to the founding of TeraDeep?
Didier Lacroix’s (the other cofounder) initial interest grew from my interest in using vision systems to share a sense of closeness with overseas parents, both being early consumers of technology from video conference cameras to skype, to more recently using multiple cameras. By tinkering with traditional vision recognition technology Didier became able to better check on his Mom. This became a larger issue as his Dad passed away and his Mom completely refused to leave her home. He was able to setup a system to track Mom as she was going through her day, ensuring that nurses, doctors and friend were helping her to maintain her routine. Didier’s own needs led him to find new technology to accelerate this process and share it with others.
In general we want machines to have similar visual abilities as us humans have, so that they can be more useful to us in our every-day life. Machines that can watch our home and family when we are away, drive us places, recognize us and respond to our needs and customs, recognize their environments and be able to automatically respond, with less and less need for human intervention, coaching, setup.
What do you feel are the leading factors enabling recent advancements in deep learning?
It always starts with good people and their innate passion to solve difficult problems! The technology is just a tool to fulfil their quest! We have been passionate about neural networks even before it was called DeepLearning! Our aim was to take inspiration from the human brain and its capabilities to deliver the next generation of devices that are better attuned to human needs and senses. And the recent surge of data such as images, text, speech enabled by cellular phones and mobile devices has created a need to understand this complex data that was not machine understandable and searchable.
And we are at the right time: the scaling up of computational resources allowed algorithms such as neural network, combined with the large availability of data, to produce significant steps forward in machine understanding of this complex data. Marrying our passion for neural networks, emulating human visual capabilities, and the grand challenge of datad-eluge — that is how TeraDeep was born.
Which industries do you think will be disrupted by deep learning in the future?
Deep Learning changes everything. The paradigm of training versus programming is immensely powerful. We are now about seeing systems that enable subject matter experts to actually train (NOT program) automated systems that generate working programs embedding specialized knowledge. Think home automation, internet of things, video understanding, security, autonomous vehicles, industrial process control and supervision, diagnostic application for medical, agricultural, remote training, etc. But the real win is uncorking the bottled up knowledge of people that have neither the background nor the ambition to become programmers. That is our goal!
What is currently being developed in your field that will be essential to future progress?
Taking Deep Learning and bringing it to the edge of the cloud, where everybody can then benefit from the technology advancements in their daily untethered life! This ambition is reflected in our company mantra “from embedded to the cloud".
How do you see computer vision having an impact on other industries?
Not just industries, the impact we aim for is to impact everyone’s life.
Which areas do you feel could benefit from cross-industry collaboration?
We need to go beyond the traditional techy approach to integrate other dimensions such as knowledge sharing and reuse, publication /monetization model, etc.
What developments can we expect to see in deep learning in the next 5 years?
An explosion of new applications launched by experts and led by the harvesting of new data and shared knowledge.
What advancements excite you most in this field?
Applying Deep Learning to make shared knowledge (from EMT life saving techniques .... to debugging how to fix a leaky toilet) more directly available to everybody. And applying it so that machines can truly understand the environment and be useful for us humans!
What do you see in the future for TeraDeep?
We have been working on demonstrating some benefits of our technology for home automation applications. This is just a start! Our passion is to launch a platform for others to bring in their own ideas, experience, excitement and start cranking new applications that can enhance your senses.
The Deep Learning Summit is taking place in San Francisco on 29-30 January. You can get more information and register here.