MIT’s Computer Science & Artificial Intelligence Lab (CSAIL) has come up with a technology called “Semantic Soft Segmentation” that uses AI to automatically separate objects in an image. It allows for easy image editing such as changing the background or compositing the figure into an entirely new scene.
The process starts with a neural network estimating the regions and features of the image. It then detects “soft transitions” such as hair and grass which in the past had to be separated by hand using programs like Photoshop. Pixels across the image are then associated with each other by color. This info is combined with the features detected by the neural network to estimate image layers, after which, the image is ready for editing. Check out the examples below.
1.
2.
3.
4.
Watch the technology in action below:
00:07 – Semantic soft segmentation explained
00:24 – How the AI works
00:44 – More examples
Share this post with a designer friend and voice your views in the comments below.