The new MIT lA can replace the background of any image

CSAIL, MIT’s research laboratory, used the potential of Automated Learning to develop a powerful image editing tool.

lA, The new MIT lA can replace the background of any image, Optocrypto

Selecting objects within an image requires a tedious process and great editing skills because many elements overlap or are part of different parts. However, this tool manages to automate the process by dividing all elements into segments and simplifying the selection of individual elements and the background.

For this purpose, the AI was trained to analyze the various characteristics of the image and to identify the characteristics of the individual objects. We can see the process he proposes in the following video sharing MIT:

As we can see, this dynamic opens up a number of possibilities, because with the perfect selection of the individual picture elements you can combine backgrounds, change scenarios, play with filters, etc…

And where editing is more complex and results on a cinematic level are required, more realistic results can be achieved because they take into account the entire context of the image to apply the corresponding changes.

We can see the technical details of this MIT project in the document they publish in this link.