Video Prediction with Appearance and Motion Conditions
This work addresses the problem of generating realistic future video frames for applications like facial expression and human action prediction, but it is incremental as it builds on existing conditional GAN methods.
The paper tackles the challenge of future uncertainty in video prediction by proposing an Appearance-Motion Conditional GAN that uses appearance and motion conditions to specify future frames, reporting favorable results on facial expression and human action datasets.
Video prediction aims to generate realistic future frames by learning dynamic visual patterns. One fundamental challenge is to deal with future uncertainty: How should a model behave when there are multiple correct, equally probable future? We propose an Appearance-Motion Conditional GAN to address this challenge. We provide appearance and motion information as conditions that specify how the future may look like, reducing the level of uncertainty. Our model consists of a generator, two discriminators taking charge of appearance and motion pathways, and a perceptual ranking module that encourages videos of similar conditions to look similar. To train our model, we develop a novel conditioning scheme that consists of different combinations of appearance and motion conditions. We evaluate our model using facial expression and human action datasets and report favorable results compared to existing methods.