AI Paper Atlas

Research map / Detection and segmentation / Weakly supervised detection and segmentation

SideInfNet: A Deep Neural Network for Semi-Automatic Semantic Segmentation with Side Information

Jing Yu Koh, Duc Thanh Nguyen, Quang-Trung Truong, Sai-Kit Yeung, Alexander Binder

ECCV 2020 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Weakly supervised semantic segmentation · 6 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
add side information (street photos/brush strokes): SideInfNet vs Deeplab-ResNetSideInfNet · Zone segmentationpixel accuracy: 64.19% -> 74.00%Improved
add side information: SideInfNet vs Deeplab-ResNetSideInfNet · BACHmIOU: 27.86% -> 47.24%Improved
add side information: SideInfNet vs Deeplab-ResNetSideInfNet · Zurich SummermIOU: 42.95% -> 58.31%Improved
add fractionally-strided convolutions to diffuse side information (Diffused vs Geotagged) (ablation)SideInfNet · Zone segmentationpixel accuracy: 64.41% -> 73.51%Improved
add adaptive inference gates, rate 0.6 (vs Diffused) (ablation)SideInfNet · Zone segmentationpixel accuracy: 73.51% -> 74.00%Improved
lower gate rate from 0.6 to 0.4 (ablation)SideInfNet · Zone segmentationpixel accuracy: 74.00% -> 73.51%Hurt
raise side information feature dimension from 64 to 128 (ablation)SideInfNet · Zurich SummermIOU: 58.31% -> 51.69%Hurt
raise side information feature dimension from 64 to 512 (ablation)SideInfNet · Zurich SummermIOU: 58.31% -> 41.37%Hurt
raise side information feature dimension above 64 (ablation)SideInfNet · BACHmIOU: 47.24% -> -Hurt
raise side information feature dimension from 64 to 128/256/512 (ablation)SideInfNet · Zone segmentationmIOU: 47.29% -> 42.11%No clear effect
use 20% instead of 0% of side information at test time (ablation)SideInfNet · Zone segmentationmIOU: 28.11% -> 34.04%Improved
use 100% instead of 0% of side information at test time (ablation)SideInfNet · BACHmIOU: 23.86% -> 47.24%Improved

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