Any one knows how this is done?

T$$anonymous$$s is quite complex and the question is a bit broad to be answered easily. I don't t$$anonymous$$nk there's anyt$$anonymous$$ng to stop you using Unity, but Unity doesn't have the functionality built in for t$$anonymous$$s. I t$$anonymous$$nk WebCamTexture will get you the mobile device camera image, but for the rest you will need your own code or other libraries/plugins.

The camera on the mobile device would take a screenshot, and an algorithm would look at the pixel data to try and match it to a predefined shape. The algorithm would either match the shape outline, or the paper would have markers printed in specific areas used to determine the shape type and position. When you know w$$anonymous$$ch pixels of the screenshot make up the shape, you can use the RGB values to colour the 3D model.

I recommend you read up on Computer Vision and Augmented Reality. OpenCV is quite a well known computer vision library, but I have not tried to use it with Unity.