AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Graph-BAS3Net: Boundary-Aware Semi-Supervised Segmentation Network With Bilateral Graph Convolution

Huimin Huang, Lanfen Lin, Yue Zhang, Yingying Xu, Jing Zheng, XiongWei Mao, Xiaohan Qian, Zhiyi Peng, Jianying Zhou, Yen-Wei Chen, Ruofeng Tong

ICCV 2021 · Label-efficient and robust learning › Semi-supervised learning · approach: Semi-supervised medical segmentation

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 boundary detection network (ablation)GN_s · LiTS 10% labeledDice (%): 83.87 -> 86.43Improved
feed boundary result to discriminator (ablation)BAS3Net · LiTS 10% labeledDice (%): 89.56 -> 91.11Improved
add semantic graph with segmentation attention map (Gs) (ablation)BAS3Net · LiTS 10% labeledDice (%): 91.11 -> 92.56Improved
add boundary graph with edge attention map (Gb) (ablation)BAS3Net · LiTS 10% labeledDice (%): 91.11 -> 91.82Improved
add bilateral (inter-task) graph reasoning (ablation)BAS3Net + BGCM · LiTS 10% labeledDice (%): 91.82 -> 93.19Improved
add self-taught semi-supervised loss on unlabeled data (ablation)Graph-BAS3Net · LiTS 10% labeledDice (%): 89.54 -> 92.99Improved
raise confidence threshold T_semi from 0.3 to 1.0 (ablation)Graph-BAS3Net · LiTS 10% labeledDice (%): 93.19 -> 91.81Hurt
increase boundary loss weight lambda_det from 0 to 1.0 (ablation)Graph-BAS3Net · LiTS 100% labeledDice (%): 93.54 -> 94.73Improved
add adversarial loss weight lambda_adv 0.005 (ablation)Graph-BAS3Net · LiTS 100% labeledDice (%): 94.73 -> 95.58Improved
increase lambda_adv from 0.005 to 0.05 (ablation)Graph-BAS3Net · LiTS 100% labeledDice (%): 95.58 -> 95.03Hurt
replace Graph-BAS3Net by BAS3Net (remove BGCM) (ablation)Graph-BAS3Net · LiTS 10% labeledDice (%): 91.11 -> 93.19Improved
add BGCM, gain at full labelsBAS3Net · LiTS 100% labeledDice (%): 95.23 -> 95.58Improved

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