Making masking easy: New Adobe Research paper showcases impressive neural network technology
posted Friday, March 24, 2017 at 6:00 AM EDT
While there are plenty of ways to make selections in Photoshop, including a variety of automatic and semi-automatic methods, complex selections remain tricky and time-consuming. Wouldn’t it be great if Adobe’s automatic tools could make traditionally difficult selections simple and easy? If Adobe Research’s latest work is any indication, the future of simple selections might very well be neural networks.
The team fed nearly 50,000 images into neural networks and their algorithm certainly learned a thing or two. Check out the two images below, drawn from a set of 1,000 test images the team used. On the far left is the original image, the second and third images use two competing existing selection methods and finally, the far right image shows the mask created by Adobe Research’s new algorithm.
The results are clearly impressive and represent a big leap in masking technology. Hopefully we will see the results of this research influence future Photoshop updates! To read all about the research, see the full paper.