AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Online Clustered Codebook

Chuanxia Zheng, Andrea Vedaldi

ICCV 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 61 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace VQ-VAE quantiser with CVQ-VAE that resamples dead codevectors from running-average anchorsVQ-VAE · CIFAR10 reconstructionrFID: 39.67 -> 24.73Improved
replace quantiser in VQGAN with CVQ-VAEVQGAN · ImageNet reconstructionrFID: 7.94 -> 1.57; usage 44% -> 100%Improved
use online running-average anchor initialization instead of offline one-time initialization (ablation)CVQ-VAE · CIFAR10 reconstructionrFID: 31.10 -> 26.62Improved
add offline anchor initialization (ablation)VQ-VAE with cosine distance · CIFAR10 reconstructionrFID: 35.14 -> 31.10Improved
add cosine distance for codebook lookup (ablation)VQ-VAE · CIFAR10 reconstructionrFID: 39.67 -> 35.14Improved
add contrastive loss to online anchor initialization (ablation)CVQ-VAE · CIFAR10 reconstructionrFID: 26.62 -> 24.73Improved
add contrastive loss on CIFAR10 SSIM (ablation)CVQ-VAE · CIFAR10 reconstructionSSIM: 0.8991 -> 0.8978No clear effect
running-average reinitialization instead of single hard feature reset (ablation)codebook reset · FFHQ reconstructionrFID: 3.91 -> 2.94Improved
vary anchor sampling method in online version (ablation)CVQ-VAE online · MNIST and CIFAR10rFID: 2.23 to 2.59 MNIST; 25.99 to 26.62 CIFAR10No clear effect
increase perplexity toward uniform codebook usage (ablation)quantisers · reconstructionrFIDNo clear effect
increase codebook dimensionality (ablation)CVQ-VAE · reconstructionreconstruction qualityMixed
use CVQ-VAE quantiser inside LDM-8LDM-8 · LSUN BedroomsFID: 3.57 -> 3.02Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy