Unsupervised Learning of Disentangled Representations from Video
This addresses the challenge of unsupervised representation learning for video analysis, with potential applications in prediction tasks, though it appears incremental in advancing disentanglement methods.
The paper tackles the problem of learning disentangled image representations from video by introducing DrNET, which factorizes frames into stationary and temporally varying components, enabling coherent future frame prediction for hundreds of steps.
We present a new model DrNET that learns disentangled image representations from video. Our approach leverages the temporal coherence of video and a novel adversarial loss to learn a representation that factorizes each frame into a stationary part and a temporally varying component. The disentangled representation can be used for a range of tasks. For example, applying a standard LSTM to the time-vary components enables prediction of future frames. We evaluate our approach on a range of synthetic and real videos, demonstrating the ability to coherently generate hundreds of steps into the future.