AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Locally Hierarchical Auto-Regressive Modeling for Image Generation

Tackgeun You, Saehoon Kim, Chiheon Kim, Doyup Lee, Bohyung Han

NeurIPS 2022 · Image generation › Autoregressive and latent generation · approach: Hierarchical VQ-VAE generation · 2 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
PixelUnshuffle/shuffle resizing instead of average pooling (ablation)HQ-VAE · ImageNet valrFID: 3.67 -> 3.14Improved
train HQ-VAE for 50 instead of 15 epochs (ablation)HQ-VAE · ImageNet valrFID: 3.14 -> 2.61Improved
locally sequential instead of global conditioning (ablation)HQ-Transformer · ImageNetFID: 23.25 -> 11.33Improved
locally hierarchical instead of locally sequential decoding (ablation)HQ-Transformer · ImageNetFID: 11.33 -> 10.01Improved
soft labels with temperature tau=1 instead of one-hot (ablation)HQ-Transformer · ImageNetFID: 10.01 -> 9.36Improved
add third code level (4x4+8x8 to 4x4+8x8+16x16) (ablation)HQ-TVAE · ImageNet valFID: 37.59 -> 15.66Improved
add third code level (8x8+16x16 to 8x8+16x16+32x32) (ablation)HQ-TVAE · ImageNet valFID: 13.59 -> 11.10Improved
add third level 4x4 to 8x8+16x16 codes (ablation)HQ-VAE · ImageNet valrFID: 18.65 -> 2.48Improved
increase HQ-VAE resolution from 8x8+16x16 to 16x16+32x32HQ-VAE · ImageNet valrFID: 2.61 -> 0.57Improved
scale HQ-TVAE from S to LHQ-TVAE · ImageNet class-conditionalFID: 9.36 -> 7.15Improved
scale HQ-TVAE from S to LHQ-TVAE · ImageNet class-conditionalthroughput: 78.87 -> 49.44Hurt
two-level hierarchical codes instead of RQ-Transformer residual codesHQ-TVAE (S) · CC3Mthroughput: 46.47 -> 54.88Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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