Facebook has released the source code of its major artificial intelligence (AI) tools that enable image segmentation and label object masks in photographs. The new move comes with an aim to advance the field of machine vision and ultimately give eyes to your computer to let it view the world just like you do.
Called DeepMask, SharpMask and MultiPathNet, the three tools by Facebook provide new algorithms to upgrade the existing machine vision. While DeepMask generates initial object masks from images, SharpMask refines those masks using lower-level features of the objects appeared. MultiPathNet, on the other hand, identifies the objects delineated by each to label each object masks and classifies the object type such as a person, dog or sheep.
The open source availability of the new tools enables developers to create new experiences for the end users using Facebook’s AI efforts. This would be a step ahead of simply tagging a person in an image. Moreover, the Mark Zuckerberg-led company might help the open source community give a machine-powered vision to visually-impaired users.
“Building off this existing computer vision technology and enabling computers to recognise objects in photos, for instance, it will be easier to search for specific images without an explicit tag on each photo. People with vision loss, too, will be able to understand what is in a photo their friends share because the system will be able to tell them, regardless of the caption posted alongside the image,” FAIR research scientist Piotr Dollar wrote in a detailed blog post.
This is not the first time when Facebook uses the open source way for its AI developments. The company released the code of its Torchnet toolkit back in June to offer some new experiences to deep learning researchers and developers.
In addition to Facebook’s progress, Microsoft and Google are also developing their native technologies to advance the machine vision field. However, making open source its exclusive developments would give an edge to the social network to grow further in the market.