Microsoft has made some strides on that front, but the company thinks that they could do better. How so? By making object recognition more contextual and more personal to the user using it. For example, object recognition might be able to recognize objects like keys, glasses, a remote control, but are these objects important to the person? What if a person who is visually impaired is trying to find their cane? What if someone moved the user’s keys to a different spot?
According to Microsoft, “There has been no effort to collect images of objects that may be particularly important to users who are blind or low vision. Providing a larger dataset for researchers and developers to use to build better AI systems could be a game changer in this area, for people who are blind or low vision in particular but also for everyone.”
To that end, the company has conducted a pilot study where they want to see and understand how those who are visually impaired might take videos, and from there, they will investigate these videos to understand what kind of objects might be of greater importance to those who are visually impaired. This in turn could then be applied to existing tech like Microsoft’s Seeing AI or even wearables like Project Tokyo.