Towards Automatic Image Editing: Learning to See another You
Xu Jia, Amir Ghodrati, Marco Pedersoli, Tinne Tuytelaars · 2016
In this paper we propose a method that aims at automatically editing an image by altering its attributes.More specifically, given an image of a certain class (e.g. a human face), the method should generate a new image as similar as possible to the given one, but with an altered visual attribute (e.g. the same face with a new pose or a different illumination).To this end, we propose a solution following an encoder-decoder pipeline.The desired attribute and the input image are independently encoded into a convolutional network and fused at feature map level.A convolutional decoder is then used to generate the target image.The result is further refined with another convolutional encoder-decoder network with the initial result and the original image as inputs.We evaluate the proposed method on MultiPIE dataset for three sub-tasks, that is, rotating faces, changing illumination and image inpainting.We show that the method is able to generate realistic images for the three tasks.