I proposed a novel generative adversarial network for facial age transformation, i.e, giving a image to the well-trained model, which will generate images that look like old people or children. It can generate high-resolution images (1024 x 1024 pixel) and very impress detailed of age features.

This network can generate very natural facial attribute during the age transfer process. When the synthetic face get older, the wrinkles become deeper and the hair becomes gray. When the synthetic face get younger, the wrinkles become smoother. When the target age is a child, it will generate the face of child and the teeth become irregular.

You could click the picture below and zoom in to see the details of the images. Each figure contains a real image, and the others are synthetic images. The last three figures are high-resolution data, each figure is around 10 MB, so it would take some time to load.

This work is made for its own sake. Inspired by Godzilla, this robot started from scratch, including skeleton design, mechanism design, circuit design and motion programming. I use Arduino to control the movement of the dinosaur by the servo motors. The pictures below is the detail of the dinosaur. You could click the first one to watch the action video.