Computers that see objects with touch
13 July 2015 Last updated at 10:01 BST
After many years of research, computers still struggle with recognising objects automatically.
Scientists at Oxford University, working in partnership with Microsoft Research, have developed Semantic Paint which allows users to 3D scan their environment and then interactively label objects.
Researchers hope that this human input will help systems of the future, such as augmented reality glasses or robot carers, to perform tasks more successfully.
BBC Click's Talia Franco talked to Prof Philip Torr, one of the lead researchers on the project.