Neural Network Checks Restaurant Entrances in Russia and Finds Improved Accessibility for Wheelchair Users
AI analysis of restaurant entrances in Novosibirsk has identified dozens of venues accessible to wheelchair users.
Artificial intelligence has helped the mapping service Yandex Maps reassess the accessibility of restaurants and cafés in Novosibirsk, identifying dozens of establishments that can be entered by people using wheelchairs.
Ramps Spotted in Photos
The Yandex Maps service has launched a computer vision system to analyze entrance areas of food-service venues. The algorithms study photographs and determine whether a person using a wheelchair can enter without obstacles.
After processing the images, the number of restaurants and cafés marked “Accessible Entrance” on Yandex Maps increased by 44 percent. Previously, the system listed 394 such locations. After the review, that number rose to 568.
The algorithms look for specific infrastructure features. The system recognizes ramps, handrails, steps, and the width of entryways. These signals are used to generate an accessibility rating for each entrance.
AI Tandem
Two neural network models are used in the project. The first analyzes images and identifies environmental elements in photographs of entrances.
The second model compares the detected elements and produces the final accessibility assessment for each venue. The technology is based on Alice AI VLM and Alice AI LLM models.
If the system cannot confidently identify an object, a human reviewer steps in. This approach minimizes the risk of errors.
Navigation for People With Limited Mobility
The updated data is already visible in Yandex Maps. Users can see the “accessible entrance” label in advance and choose a suitable venue.
According to project specialists, the initiative helps people with limited mobility navigate urban environments more easily. The digital analysis of photographs is gradually enriching the city map with information about the accessibility of public spaces.
