CrevNet: Conditionally Reversible Video Prediction
This addresses memory limitations in video prediction models, enabling broader application scenarios, though it appears incremental as it builds on reversible architectures.
The paper tackles the high memory consumption of resolution-preserving blocks in video prediction by proposing CrevNet, a conditionally reversible network that uses reversible architectures to build a bijective two-way autoencoder and recurrent predictor, resulting in theoretically guaranteed no information loss, much lower memory consumption, and computational efficiency.
Applying resolution-preserving blocks is a common practice to maximize information preservation in video prediction, yet their high memory consumption greatly limits their application scenarios. We propose CrevNet, a Conditionally Reversible Network that uses reversible architectures to build a bijective two-way autoencoder and its complementary recurrent predictor. Our model enjoys the theoretically guaranteed property of no information loss during the feature extraction, much lower memory consumption and computational efficiency.