Dual-Motion Transfer GAN

Generating videos with content and motion variations is a challenging task in computer vision. The proposed model is trained in an end-to-end manner, without the need to utilize pre-defined motion features like pose or facial landmarks.