Using a frozen learned-compression encoder as the ViT patch embedder yields a 4x token reduction and 63% FLOP savings, with accuracy gains shown only on one small dataset.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Compress image to patches for Vision Transformer
Using a frozen learned-compression encoder as the ViT patch embedder yields a 4x token reduction and 63% FLOP savings, with accuracy gains shown only on one small dataset.