activity
20162025
most citedSoft-to-Hard Vector Quantization for End-to-End Learning Compressible Representations

355 citations · 1.3k across the 14 of their papers we have counts for

collaborators
Showing 2023 · cs.CVShow all

2 papers · 2 filters

cs.CV2023

GIVT: Generative Infinite-Vocabulary Transformers

Michael Tschannen, Cian Eastwood, Fabian Mentzer

We introduce Generative Infinite-Vocabulary Transformers (GIVT) which generate vector sequences with real-valued entries, instead of discrete tokens from a finite vocabulary. To th…

cs.CV2023★ 13 cited

Finite Scalar Quantization: VQ-VAE Made Simple

Fabian Mentzer, David Minnen, Eirikur Agustsson +1

We propose to replace vector quantization (VQ) in the latent representation of VQ-VAEs with a simple scheme termed finite scalar quantization (FSQ), where we project the VAE repres…