CSCI 599 FINAL PROJECT
Human Face to Human Face
We achieve this by introducing two novel sub-networks. The first Land- mark Disentangling Network (LD-Net) learns to disentangle identity from head poses/expression, predicting facial landmarks that preserve the identity of the target while combining expressions and poses from another driving subject. The second Feature Dictionary-based Generative Adversarial Network (FD-GAN) learns to transform landmark positions into a personalized video portrait of someone given a single target image, which allows subject agnostic reenactment of a portrait that preserves a recognizable identity of the target and can be applied to unseen identities without subject-specific training.
Human Face to Game Character in Unity
The general idea of this method is to build a mapping function that translates human face to game character face. An end to end solution is to build an image-to-image translation generative network whose input is human face and output is game character face. But getting pairs of images of human face and game character face is hard, and we need to build a special dataset for each person even if we do not change the target game character. So we propose another solution.
Human Face to Game Character (Machine Learning)
The general idea of this method is to build a mapping function that translates human face to game character face. An end to end solution is to build an image-to-image translation generative network whose input is human face and output is game character face. But getting pairs of images of human face and game character face is hard, and we need to build a special dataset for each person even if we do not change the target game character. So we propose another solution.