Improving Road Safety Using Artificial Intelligence
When it comes to determining how safe it is to drive in any given country, you can read all the statistics you want, but the best way to get some ground truth is to rent a car and see for yourself. For example, take the country of Rwanda. It’s the second most dangerous country in Africa to drive, and a place where a sophisticated series of hand gestures are used to warn oncoming drivers about traffic police. Everybody just learns to skirt the system, and driving safely just becomes an impediment to getting somewhere quicker. Should you actually be stopped by the police, 20,000 Rwandan francs (about $20) will do the trick, so the purpose of having traffic police is just to extort a form of road tax. There are many countries in the world where people are comfortable taking risks while driving, and one of them is the great Kingdom of Saudi Arabia.
The second leading cause of death in Saudi Arabia is getting killed in an auto accident. Most of the people we spoke with in Riyadh speculated that the problem results from the large number of drivers who hail from other countries where driving isn’t exactly that safe. The end result is a melting pot of drivers on the road, each with their own set of road rules. Regardless of the reason why people are comfortable taking more risks when driving, spend some time in the Kingdom and you’ll quickly realize two things – Saudi people are incredibly friendly, and road safety is a problem that needs to be dealt with. One Saudi Arabian startup – Hazen.ai – is looking to improve road safety around the globe using artificial intelligence.
Founded in 2018 and based out of Makkah, Saudi Arabia, Hazen.ai has taken in an undisclosed amount of funding (rumor has it their seed round made seven figures) from Saudi Aramco’s venture capital arm, Wa’ed Ventures, to develop computer vision solutions for traffic – “smart cameras for a smarter city.” We sat down to speak with co-founder and CEO of Hazen.ai, Sohaib Khan, who worked as a professor of computer vision for fifteen years before deciding to embark on an entrepreneurial quest to save lives.
Today, traffic enforcement is conducted using expensive equipment like LiDAR and radar devices which capture speeding or red-light infractions. Tomorrow, it just might be computer vision algorithms. Turns out that Saudi Arabia, and the region as a whole, is the perfect place to build such a solution for any number of reasons.
Why Saudi Arabia?
Tech news is inundated with stories of startups trying to make cars drive themselves, or cars equipped with cameras that assess other people’s driving capabilities, or even cameras facing inwards which detect driver fatigue or a lack of concentration. What seems to be lacking is a technology solution that can detect moving violations from a stationary position. Here are some examples of moving violations that you might want to detect with such a solution.
The environment in the region is so rich with traffic pattern variations that traffic management companies from all over the world come to Saudi Arabia to set up shop. A rich variety of edge cases for training algorithms means that if algorithms work well here, they’ll work well anywhere. Improving traffic safety is a problem that governments are more than willing to spend the money needed to fix, and there aren’t a whole lot of organized groups that would oppose the deployment of safety mechanisms. It’s the perfect place to build a solution and then sell it across the globe, something that Hazen.ai is actively doing.
The Hazen.ai Platform
We’ve talked a lot before about training data that’s being used to train computer vision algorithms for autonomous vehicles. What you may not be aware of is that most of this data is based on Western markets where things operate a whole lot differently. Hazen.ai seeded their computer vision algorithms using Western data sets, then started using customer data from various geographies to fine-tune their models (what’s called model adaptation). Figuring out what entails a “moving violation” will differ based on what country the camera is placed, or even what physical location it occupies. The core technology behind the Hazen.ai solution is the ability to classify vehicles in any environment or lighting configuration, then determine the vehicles’ trajectory.
Once trained, these algorithms are then placed “at the edge” where a credit card-sized GPU chipset from NVIDIA resides in a small box that plugs into any IP-enabled camera that is conducting surveillance, like CCTV cameras, for example. The camera’s feed is then redirected into the box where moving violations can be assessed in milliseconds. Once a violation has been identified, a request is made to an automatic license plate reader (ALPR) to identify the vehicle. That identification image along with video evidence of the infraction is then assembled as a “violation package” that’s configured based on the legal requirements of the country their client resides in. It’s just what traffic enforcement officers need to issue a citation and curb moving violations without having to deploy more bobbies on the beat.
The Way Forward
Mr. Khan spoke proudly of the prestigious product innovation award his startup managed to land at Gulf Traffic 2018 in Dubai. At that time, they were the only company present dabbling in computer vision for traffic safety. When they attended that same event in December 2019, that all changed, with numerous companies beginning to tackle this problem. Hazen.ai has a good head start, having landed their first commercial client from a developed market, along with plenty of potential clients they’re doing pilot projects with. In one example, Hazen.ai was given 24 hours of traffic data to classify infractions on. When all was said and done, they discovered that drivers had actually figured out ways to cheat the existing traffic cameras – the equivalent of a head fake for all you basketball players out there. They’re also doing classification work in other countries like Pakistan where you’re liable to see bikes, motorcycles, tuk-tuks, buses, people, and just about everything else on the road.
There are plenty of other use cases that Hazen.ai might address such as traffic management or parking, but moving violations are where their bread and butter is at right now. Insurance companies are likely to find this technology interesting, as are governments, but Mr. Khan thinks the way forward is to target bigger traffic management companies that are just starting to explore this space. As a small startup, Hazen.ai can deploy their solution much quicker by partnering with companies that already have cameras or ALPR devices deployed on street corners. More importantly, their core technology – vehicle classification algorithms – can be used for autonomous driving when the time comes. If you’re a company that already mastered autonomous driving for Western markets, you just might need some data from places where things are a bit more heterogeneous. Anyone who is able to amass such a big data set will be sitting pretty when that time comes.
Pundits are always quick to call for the death of oil in the face of electric cars, not realizing just how far away that point is for the rest of the world, which doesn’t reside in developed markets like ‘Murica. The same can be said for self-driving vehicles. Until we reach a point in time where all the world’s nations can afford to upgrade to green transportation methods, there will be traffic safety issues to address because understandably, all cultures choose to drive differently depending on their norms. It’s difficult to change driving behavior without using a carrot or a stick. In countries where auto insurance isn’t so popular, the carrot won’t work. This means that solutions like the one being built by Hazen.ai can save lots of lives while we wait for the world to catch up to autonomy, and perhaps make some investors a whole lot of money in the process.
The only issue with exciting tech startups is that retail investors cannot invest in them. This is why we created “The Nanalyze Disruptive Tech Portfolio Report,” a portfolio of more than 20 disruptive tech stocks we love so much we’ve invested in them ourselves. We carefully reviewed the hundreds of stocks and dozen or so ETFs we’ve ever written about and rated each of them. Find out which tech stocks we love, like, and avoid in this report, now available for all Nanalyze Premium annual subscribers.