InMoDeGAN: Interpretable Motion Decomposition Generative Adversarial Network for Video Generation (original) (raw)

[PDF] [Code]


Abstract


1. Random Generation

We randomly sample different appearance noises za and motion noise sequences {zm0,zm1,...,zmT-1} on VoxCeleb2-mini in two resolutions (128x128, 64x64) and BAIR-robot.

5. Controllable Generation