VJ-VCR applies variance-covariance regularization to a video joint-embedding predictive architecture and beats a generative baseline at probing dynamics from frozen representations.
Each video contains 300 frames at 24 frames per second at 320x240 resolution
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Video Representation Learning with Joint-Embedding Predictive Architectures
VJ-VCR applies variance-covariance regularization to a video joint-embedding predictive architecture and beats a generative baseline at probing dynamics from frozen representations.