Canadian startup MicroTraffic follows four basic steps to reduce traffic risks and create safer routes: Collect data, analyze them, make a diagnosis and propose a solution. This process allows the company to create safety plans capable of reducing traffic accidents by 80%. And it led her to win first place in the “Smart Cities, Transportation & Logistics” category of the SXSW Pitch, held in March 2021. Founded in 2017, MicroTraffic helps cities to plan improvements in traffic safety and monitor the effectiveness of the implemented changes, with an AI-powered Computer Vision solution. “The results of the prototype helped so much in understanding the risks and the appropriate countermeasures, that we decided to transform the consulting project into the MicroTraffic monitoring solution”, explains the Road Safety Engineer and company CEO, Craig Milligan. Last year, the startup added 25 cities to its pilot program and its goal is to be present in a total of 170 municipalities next year. The company has already doubled the size of its computer vision team and tripled the size of its road safety team. In this interview, Milligan comments on how predictive techniques are being used for road safety analysis. MicroTraffic was created by engineers with extensive experience in diagnostic studies. Our concern was that historical accident data, which are often used in these studies, do not reveal latent risk factors. Consequently, they are an inadequate diagnostic tool. Using traditional techniques, a city’s traffic engineers only discover a risk after a fatal accident has occurred, even if accidents occur almost every day on that spot. With our street monitoring solution, engineers can record these “near misses”, understand the risk factors before an incident even occurs and correct the problem. The data that excites me the most are those that show a reduction in near misses at an intersection. Being able to reduce from 80% to 90% of the risk at an intersection is what makes us wake up in the morning and this is happening more and more. This is our mission, a vision of a world free from fatal traffic accidents, thanks to technology. MicroTraffic's work begins with the collection of images from the cameras present in the locations to be analyzed. Many cities invest in camera systems or a closed-circuit television (CCTV) system. What we do is turn that data into useful information. Our computer vision algorithm uses Deep Learning to detect all track users in each video frame. Then, it creates a link between successive frames to obtain a path. We still have a spatial location algorithm that translates the video location into terrestrial coordinates and creates what we call a trajectory database. This permanently updated database represents the location and time of each user on that road. And it allows us to dynamically understand the relationship between users and compare different trajectories. The algorithm processes pairs of trajectories to assess risk, taking into account speed, spatial proximity, angular relationships between road users, and how vulnerable they are. Based on these factors, we categorize the interaction at a given level of risk, from a benign event to a critical risk interaction. A latent risk factor, for example, is a risk that has not yet been expressed in the crash data, but which has the potential to result in serious injury or fatalities. With this work, we find out precisely which parts of an intersection are mostly at risk and which road users are typically in a risky situation at that location. Then, risks are reported for very specific areas of that intersection. Therefore, raw data produced by the cameras, which are commonly neglected, can become an engineering or intervention project on the analyzed street. From the diagnosis of the problem, we can understand how to change the physical layout of intersections, better accommodate road users, and separate conflicting routes with the objective of reducing risks. Recording videos is an obstacle. There are no cameras at all intersections and sometimes the cameras may even exist, but the design of the CCTV network does not allow us to do much with them. In some cities, cameras are frequently offline and do not record what is necessary for us to make a diagnosis. One of the biggest challenges for cities that want to use our system is to have a good CCTV system. Yes, the connection can also be an obstacle, but I would characterize it as part of the CCTV system. When we do video analysis, there is a computational and bandwidth dependency. But it is possible to choose between applying a lot of computational power at the edge and not worrying about bandwidth or keeping this computational power centralized and having a greater demand for bandwidth. We use centralized computing and, in the future, we will have an edge option. Some cities will have to use our edge solution, but when doing so, there will be a need to put more computing power on the edge, which will require a more advanced camera. Today MicroTraffic is able to create a pilot in any municipality, regardless of the technological situation. As we want to demonstrate the value of our technology, we can install temporary cameras, and remove them after the data captured. So it is possible to test the solution. If the city likes the technology and is building a smart city monitoring structure, it is possible to invest in that technology. Some municipalities do not plan to make a large investment in bandwidth, so they have been using temporary cameras for years. In such cases, the devices are placed at an intersection, the risk factors are discovered and treated, the risk reduction is measured and the device is installed elsewhere. MicroTraffic tries not to restrict itself to locations that already have robust IoT monitoring configurations. However, the locations that have this technology are at a great advantage because our solutions can be used almost instantly. Monitoring technologies, especially those aimed at security, are very important trends in smart cities. Between the 1970s and 2010, there was a big reduction in traffic accidents because we did everything that was easy to do without technology. But from 2010 on, progress in traffic safety has almost stopped, so the hope for many cities is that monitoring, IoT sensors, and artificial intelligence may unlock the next step in road safety in a variety of applications. First, we were very happy to have been selected to participate in the SXSW Pitch and then to win the prize. It was a great experience. The judges also helped a lot. The questions we were asked led us to think about important elements of our business, but I think we gave good enough answers at the time. For us, the prize means two things. First, it is a validation that what we are doing is important, exciting and that other people see value in something that we are passionate about. The other thing is that it gives us exposure. We hope that, through the award and the publicity surrounding it, more cities will come to know us.