Private Collection is a technical prototype for an app which would allow you to choose between different levels of abstraction to mask sensitive personal information encountered in augmented reality (AR) applications within the home. Technology in the home is exciting, but comes with a lot of questions about your personal security and privacy. Private Collection addresses these questions.
The information can be personalised to be anything from whole spaces, objects and other people. The idea is to showcase that treating people's data with respect does not exclude the use of intelligent AR applications in the home. In contrast, it can be leveraged to create more personal and trustful relationships between people and service providers.
'Trust is fundamentally about sharing controls between all entities involved, over time.'
For this experiment, CIRG worked with machine learning based image segmentation techniques to identify humans, pets, doors, windows, images and objects. For better visual treatment of images and videos, they used human pose estimation, allowing them to get higher accuracy detecting the nuances of the human body. They were then able to extract all of these entities from the footage and to allow users to adjust their individual abstraction levels. The visual treatment would apply a tunable filter, creating an effect that literally 'paints' over the extracted entities in a soft and silky, watercolour-like way, blending in with domestic environments.
The user can decide what types of objects or people should be automatically abstracted. By long-pressing any one of these abstracted objects or people, the user would be presented with a slider to let them manually configure the level of abstraction for that particular selection. Maybe there are things they actually want to share, but that would be up to them.
'To establish trust we need to give users the option to have more control over their data.'