Abstract

At least 2.2 billion people are visually impaired, with the main problem being a lack of autonomy and safety when moving around the city. To solve those needs, a wearable assistive system for vision impairment people, with low cost and open-source applications, integrating machine learning and deep learning was developed. We used a mixed methodology, including semi-structured interviews with visually impaired people and specialists in visually impaired needs, including them both in the design process. This allowed us to have an understatement of the specific needs and to generate improvements in the design. The results are auspicious for the feasibility of this type of development, the main considerations and lessons learned are noted, leaving some opportunities for future research.