AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Entroformer: A Transformer-based Entropy Model for Learned Image Compression

Yichen Qian, Xiuyu Sun, Ming Lin, Zhiyu Tan, Rong Jin

ICLR 2022 · Architectures and efficiency › Attention and transformers · approach: Attention head and block redesign · 14 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 the CNN entropy model with Entroformer using hyperprior and contextLearned image compression · Kodak, low bit ratesbpp reduction: 5.2%Improved
Replace BPG with EntroformerImage compression · Kodak, low bit ratesbpp reduction: 20.5%Improved
Use relative 2D diamond position encoding instead of the CNN baseline (ablation)Entroformer context model · Kodak, λ = 0.02bpp: 0.6621 -> 0.6238Improved
Replace no position encoding with absolute 2D position encoding (ablation)Transformer context model · Kodak, λ = 0.02bpp: 0.6569 -> 0.6563Improved
Use relative 1D+1D position encoding instead of absolute 2D position encoding (ablation)Transformer context model · Kodak, λ = 0.02bpp: 0.6563 -> 0.6358Improved
Extend relative position encoding from 1D+1D to 2D (ablation)Transformer context model · Kodak, λ = 0.02bpp: 0.6358 -> 0.6296Improved
Add a diamond boundary to relative 2D position encoding (ablation)Transformer context model · Kodak, λ = 0.02bpp: 0.6296 -> 0.6238Improved
Use relative instead of absolute position encoding (ablation)Entroformer · Tecnick, 300 × 300bpp: 0.7654 -> 0.7180Improved
Use relative instead of absolute position encoding (ablation)Entroformer · Tecnick, 600 × 600bpp: 0.5360 -> 0.5057Improved
Use relative instead of absolute position encoding (ablation)Entroformer · Tecnick, 900 × 900bpp: 0.4938 -> 0.4154Improved
Use relative instead of absolute position encoding (ablation)Entroformer · Tecnick, 1200 × 1200bpp: 0.4808 -> 0.4099Improved
Add bidirectional instead of unidirectional context for the second latent slice (ablation)Entroformer with hyperprior · Kodak, second latent slicebpp: 0.3105 -> 0.2972Improved

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