AI Paper Atlas

Research map / Architectures and efficiency / Quantization

CABM: Content-Aware Bit Mapping for Single Image Super-Resolution Network With Large Input

Senmao Tian, Ming Lu, Jiaming Liu, Yandong Guo, Yurong Chen, Shunli Zhang

CVPR 2023 · Architectures and efficiency › Quantization · approach: Ultra-low-bit LLM quantization · 14 citations (OpenAlex)

Read the paper

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 CADyQ's MLP bit selector with CABM edge-to-bit lookup tablesCARN · Urban100, x4BitOPs: 3.13G -> 1.57GImproved
replace CADyQ's MLP bit selector with CABM edge-to-bit lookup tablesEDSR · Urban100, x4PSNR (dB): 25.90 -> 25.95No clear effect
replace CADyQ's MLP bit selector with CABM edge-to-bit lookup tablesEDSR · Test2K, x4PSNR (dB): 27.54 -> 27.57No clear effect
replace CADyQ's MLP bit selector with CABM edge-to-bit lookup tablesEDSR · Test4K, x4BitOPs: 82.37G -> 47.70GImproved
add fine-tuning after replacing MLP-selected bits with CABM lookup tables (ablation)EDSR · Urban100, x4PSNR (dB): 25.77 -> 25.95Improved
add fine-tuning after replacing MLP-selected bits with CABM lookup tables (ablation)IDN · Urban100, x4PSNR (dB): 24.32 -> 25.58Improved
increase the calibration expansion range Δe from 10 to 80 (ablation)EDSR CABM · Urban100, x4PSNR (dB): 25.95 -> 25.87Hurt
choose maximum- rather than minimum-BitOPs configurations per interval (ablation)EDSR CABM · Urban100, x4PSNR (dB): 25.95 -> 26.04Improved
choose maximum- rather than minimum-BitOPs configurations per interval (ablation)EDSR CABM · Urban100, x4FAB: 5.90 -> 6.64Hurt
randomly choose rather than select minimum-BitOPs configurations per interval (ablation)EDSR CABM · Urban100, x4PSNR (dB): 25.95 -> 26.01Improved
sample subnet difficulty in proportion to BitOPs rather than uniformly (ablation)EDSR supernet · Urban100PSNR (dB): 25.94 -> 26.01Improved

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