Noise-based and masking-based visual pre-training work better together when corruption is inside the encoder, noise is added at lower-layer features, and masked and noised tokens are explicitly separated.
Context autoencoder for self- supervised representation learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Improving Generative Pre-Training: An In-depth Study of Masked Image Modeling and Denoising Models
Noise-based and masking-based visual pre-training work better together when corruption is inside the encoder, noise is added at lower-layer features, and masked and noised tokens are explicitly separated.