AI Paper Atlas

Research map / Detection and segmentation / CNN features

ISNet: Integrate Image-Level and Semantic-Level Context for Semantic Segmentation

Zhenchao Jin, Bin Liu, Qi Chu, Nenghai Yu

ICCV 2021 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 92 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
add the image-level context module (ILCM) (ablation)ResNet-50 baseline · ADE20K val (single-scale)mIoU: 36.96 -> 42.50Improved
add the semantic-level context module (SLCM) (ablation)ResNet-50 baseline · ADE20K val (single-scale)mIoU: 36.96 -> 42.89Improved
add SLCM on top of ILCM (ablation)ISNet ResNet-50 · ADE20K val (single-scale)mIoU: 42.50 -> 44.09Improved
add ILCM on top of SLCM (ablation)ISNet ResNet-50 · ADE20K val (single-scale)mIoU: 42.89 -> 44.09Improved
replace ILCM with ASPP as image-level moduleISNet ResNet-101 · LIP valmIoU: 55.41 -> 56.96Improved
add SLCM to ASPPASPP ResNet-101 · LIP valmIoU: 55.34 -> 56.96Improved
replace ILCM with ASPP as image-level moduleISNet ResNet-101 · Cityscapes val (single-scale)mIoU: 80.56 -> 81.10Improved
use ResNet-101 instead of ResNet-50ISNet · ADE20K val (multi-scale + flip)mIoU: 45.04 -> 47.31Improved
use ResNeSt-101 instead of ResNet-101ISNet · ADE20K val (multi-scale + flip)mIoU: 47.31 -> 47.55Improved
use ResNeSt-101 instead of ResNet-101ISNet · COCOStuff test (multi-scale + flip)mIoU: 41.60 -> 42.08Improved
use ResNeSt-101 instead of ResNet-101ISNet · LIP valmIoU: 55.41 -> 56.81Improved
use ISNet context module instead of DNLResNet-50 context scheme · ADE20K val (single-scale)mIoU: 43.50 -> 44.09Improved

Later papers that compared with this method

7 later accepted papers report IS-Net as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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