Inclusive AI refers to the development and use of AI systems that are designed to be accessible, fair, and equitable for all individuals, regardless of race, gender, age, ability, geography, region, language, or other characteristics. This involves ensuring that AI models are not biased and that they account for and accommodate diverse user needs and experiences.

An article published by Harvard Business Review in 2020 highlights the challenges of building a fair AI and suggests three key principles for creating fair AI systems: avoiding bias, promoting transparency, and ensuring accountability. The article stresses the importance of involving diverse perspectives in the development process and continuously evaluating and refining AI systems to address potential biases and unintended consequences.

Visual semantic segmentation is at the core of our DeepWalk™ AI stack at footpath.ai. We use semantic segmentation to understand sidewalk scenes in our 360° street-level imagery such as detecting the sidewalks 🚶‍♀️ and their boundaries, crosswalks 🚸 , and other static or dynamic objects on the sidewalk. Together with depth estimation, we project the identified and detected features and objects on a map with georeferenced points, lines, and polygons.

Sidewalks and street furniture on the sidewalks can look very different in different regions. While our AI models are trained on a relatively small training data set, for now, they perform quite well in a small number of cities that we have had our focus on over the past two years. However, we still have a long way to go to achieve our long-term goal of training inclusive and fair AI models that perform equally well in every city in the world, be it the streets of Seoul, Santiago, Manila, Sydney, or London. A model that is not biased toward a specific urban form in a region or country.

Get in touch to learn how footpath.ai can help your city transform into a more walkable city with better map data. We welcome opportunities for partnership and collaboration with research institutes, corporates, and other startups in the mapping/geospatial industry.