Luxonis wants to use computer vision to protect cyclists

6 Jul 2020

Luxonis CEO Brandon Gilles. Image: Luxonis

Our Start-up of the Week is Luxonis, a Colorado-based company developing embedded artificial intelligence and computer vision technology.

Brandon Gilles is the CEO of Luxonis. An entrepreneur with a background in electrical engineering, he told Siliconrepublic.com that he is driven by the urge to make things and do things that matter to the world.

During his time studying electrical engineering, he found that modern engineering schools run pretty narrow courses that don’t often allow for a broad study of the different disciplines within the field of engineering.

“I still wanted to learn as much as I could about other disciplines in engineering,” Gilles said. “So I took mechanical engineering courses, worked in aerospace on the DANDE student satellite that was successfully launched to space, and tried to cover as many sub-fields of electrical engineering as possible.”

These subfields of engineering include radio frequency engineering, power electronics, embedded systems, software and, most recently for Gilles, artificial intelligence (AI), machine learning and computer vision.

Founding Luxonis

Gilles became particularly interested in pursuing an education in AI, machine learning and computer vision after one of his mentors left a “fantastic job” because he believed that machine learning and AI presented what could be “the biggest opportunity of his career”.

A year later, Gilles said he got “up to speed” on the topic and followed suit by leaving his job. After quitting his job, he said his plans for the company changed after a number of his friends, colleagues and acquaintances were struck by distracted drivers while cycling to and from work.

Some of his friends were left critically injured and in one case there was a fatality. Gilles sought to see if he could develop an embedded device that could detect when such collisions were going to occur to take action and prevent accidents.

“In doing so, we prototyped a solution quickly to prove the concept and, after showing that such as device is possible and indeed works well, we focused on productising,” Gilles said.

“We found that although the perfect AI and computer vision chip for such a product existed – the Myriad X – there was no platform in hardware, firmware or software to use it how we needed to use it: for real-time spatial perception.”

Gilles said: “We had a choice to make: walk away from the effort or build that platform.”

The technology

The team at Luxonis went on to build a platform called DepthAI. Gilles said that it has been open sourced and the reception from the open source community has been “fantastic”.

“The first missing piece was the hardware itself: the board designs required to build such an embedded spatial AI solution,” Gilles said. “We now have a slew covering effectively all the permutations and use cases of the Myriad X. We make a Myriad X system on module (SoM) with open-source reference designs, which made productising a breeze.”

Gilles said: “Some folks buy the reference designs directly and integrate into their own assemblies, while others use the references to design their own custom printed circuit board (PCB) and only buy the SoM from us.”

The Luxonis CEO said that the SoM works with a wealth of code that the company has open-sourced through an MIT license, which makes getting proof of concept up and running in an application take a matter of hours. He shared the repositories for Python interface and examples, C++ Core and C++ API, and online AI/ML training leveraging Google Colab.

“A good example of such application is a student at the University of Colorado whose mission is to pick strawberries autonomously,” Gilles said. “Using DepthAI’s open source stack, he was able to build an automated strawberry picker over the course of a weekend.”

Future plans

While Luxonis has built quite a platform, Gilles has stuck to his early objective of protecting cyclists. He said that bike commuting is “an idyllic form of the future” and he wants to do his part to try and build or enable that future.

“Being a technologist and a trained electrical engineer, the most probable way to help is through technology,” he said. “We’re at an inflection point now where such a system is just becoming tractable. DepthAI, powered by Intel Movidius Myriad technology, is making it tractable.”

The CEO said that Covid-19 “threw a wrench into everything” and that Luxonis saw many of its customers pull back on non-essential spending or turn to substantial lay-offs.

“Our biggest customer laid off over 700 of their 900 employees in one week alone, just to stay afloat and keep paying the remaining 200,” Gilles said. “The CEO had the choice to give up and shut the whole business down, or lay off the majority and try to keep fighting to keep 200 employed.”

Gilles said that Luxonis has let companies in these situations out of payment obligations, “given that they had to lay off workers with babies on the way”. Prior to the slowdown caused by Covid-19, the CEO said that Luxonis had orders going through “in the thousands of units”, which equalled more than the total investment in the company to that date.

“When the Covid-19 economic damage really started to hit, all of this dried up. Over the past few weeks, it has started to look very promising again and business, revenue and cashflow are moving back to break-even territory. For a nascent hardware business, we’re excited about this.”

Backlash against cycling technology

Besides a global pandemic, Gilles said that other challenges Luxonis has faced since it was founded have been the resistance that many feel towards the bicycle market and associated products.

“Most folks are very pessimistic about our mission and many are too angry at the prospect of helping people to ride bikes,” he said.

“One prominent businessman we reached out to on the idea of the product, halfway through explaining how my friends, colleagues and acquaintances had been injured or killed by distracted drivers, interrupted to tell us that people who ride bikes don’t get hit by cars – that they hit the cars.”

Gilles said that this experience was not a once-off. He said there is a strong vein of anti-biking culture in the US and that many of the people who commute by bike have their own stories of cars intentionally trying to intimidate them off of the roads.

Despite that, Gilles remains steadfast in his mission to make cycling safer and make AI and computer vision more accessible to other businesses that have an idea that they’re not sure how to pursue.

“If you’re trying to do something that matters, you’re going to find yourself in preposterous situations,” Gilles said. “Be willing to laugh it off when things go terribly wrong – and then keep going.”

Want stories like this and more direct to your inbox? Sign up for Tech Trends, Silicon Republic’s weekly digest of need-to-know tech news.

Kelly Earley was a journalist with Silicon Republic

editorial@siliconrepublic.com