Traffic Vision is developing software that uses computer vision to make roads safer by analyzing traffic footage and detecting dangerous situations.

One part of the project aims to preserve dashcam footage that is usually overwritten and use it to build the first near-miss map of Samarkand. A near-miss is measured by the time gap between a pedestrian leaving a road space and a vehicle entering it, helping identify dangerous crossings before serious crashes occur.

The project will collect one month of footage from ten dashcams installed in taxis that regularly pass through high-risk junctions. Students will label the footage, and the resulting data will be used to train and improve the detection system. The outcome will include a public map of the ten most dangerous crossings in Samarkand, together with an anonymized annotated dataset available for others to use.

The team is also developing software capable of automatically detecting road incidents and sending information to emergency response services, including the type of incident, date and time, and the coordinates where it occurred.

The project’s goal is to bridge the gap between a working technical prototype and real-world deployment, creating tools that could help prevent accidents and enable faster emergency response.