From weeds to holes: how AI helps detect what the human eye can't see

21 November, 2022


Smart System Technologies is a young Kazakhstani IT company with founder Anuar Sarsembinov that has developed two AI solutions. One of them is the Smart Road Diagnostic System, which analyzes photos of roads and finds even the most inconspicuous damage.  

— How did your company come into being?

Our story is not exactly a startup: we did not look for a problem, and we did not do any preliminary customer development. We were given a task, and we solved it. The Smart Road Diagnostic System was the first project that started our company. 

That was in 2020.  I had already graduated from university, and I had some small projects of my own. One day, I was sitting in a café with my computer, minding my own business. And then one man comes up to me. He saw that I was into programming, and asked if I could create a solution that would automatically scan the road surface.

It turns out that there's a company in Kazakhstan that takes photos of almost all the roads in the country to look for defects on them and then passes this data on to road owners. Then all these photos were manually examined by several people in each area of the country, wasting a lot of time, but still missing some cracks.

Then I remembered that I had a friend who had invented his startup: his algorithm analysed photos of fields to detect weeds and then spot-sprayed chemicals on them. I thought that if some kind of software analyses photos and looks for certain plant species, then it could be trained to look for holes and cracks in the asphalt. 

So, we teamed up with this friend and then got a couple of other guys to join the team. We were able to automate the processes: they would send us the photos, and we would upload them into the system and give an analysis in the form of an excel spreadsheet. 

— How have optimised that work?

The system is now able to detect up to 10 kinds of road defects with an accuracy of over 95 %. Moreover, we can detect as much as one and a half times as much damage in just one to two hours, and it only takes one technician.

— How does the system that takes pictures of the roads look like? Is it some kind of copter?

No, that would be too complicated. It looks like a structure with cameras that are attached to the frame of a car. 

— How scalable is your idea? If Kazakhstan has a monopoly on monitoring road quality...

We have established partnerships with Russian companies. And specifically with those who produce the scanning equipment itself. There are 2 or 3 such companies in Russia. We sold a licence to one of them to use our technology. So for each purchase of equipment from that company, we get a percentage. And this is profitable for us because the equipment needs to be replaced every 3-4 years, which means that there will always be deals. 

As for other markets, we have so far suspended the search for them because we are focusing on our other product, MultiCode.

Read more about the company's second solution in our next article.

Telegram   /   Instagram   /   VС.ru   /   LinkedIn