AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Non-uniform Step Size Quantization for Accurate Post-Training Quantization

Sangyun Oh, Hyeonuk Sim, Jounghyun Kim, Jongeun Lee

ECCV 2022 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 24 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
subset quantization instead of BRECQ weight quantization (4-bit)ResNet-18 · ImageNet W4/A32top-1 degradation (%): 0.38 -> 0.30Improved
subset quantization instead of BRECQ (3-bit weights)ResNet-50 · ImageNet W3/A32top-1 degradation (%): 1.39 -> 0.99Improved
subset quantization instead of BRECQ (2-bit weights)ResNet-18 · ImageNet W2/A32top-1 degradation (%): 4.78 -> 4.14Improved
subset quantization instead of BRECQ (2-bit weights)ResNet-50 · ImageNet W2/A32top-1 degradation (%): 4.60 -> 3.86Improved
subset quantization instead of BRECQ with 4/4 quantizationResNet-18 · ImageNet W4/A4top-1 degradation (%): 1.48 -> 1.21Improved
subset quantization instead of BRECQ with 4/4 quantizationResNet-50 · ImageNet W4/A4top-1 degradation (%): 1.95 -> 1.48Hurt
subset quantization instead of BRECQ with 3/3 quantizationResNet-18 · ImageNet W3/A3top-1 degradation (%): 5.05 -> 4.62Hurt
subset quantization instead of BRECQ with 3/3 quantizationResNet-50 · ImageNet W3/A3top-1 degradation (%): 8.04 -> 6.80Hurt
subset quantization instead of BRECQ with 2/4 quantizationResNet-18 · ImageNet W2/A4top-1 degradation (%): 6.28 -> 5.06Improved
subset quantization instead of PWLQ at 4-bit weightsDeepLabV3+ MobileNetV2 · segmentation W4/A8mIoU degradation (%): 3.15 -> 1.85Improved
subset quantization instead of uniform at 4-bit weightsDeepLabV3+ · segmentation W4/A8mIoU degradation (%): 20.76 -> 1.85Improved
subset quantization instead of uniform at 4-bit weightsSSD-Lite · detection W4/A8mAP degradation (%): 3.91 -> 0.38Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 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