AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Med-DANet: Dynamic Architecture Network for Efficient Medical Volumetric Segmentation

Wenxuan Wang, Chen Chen, Jing Wang, Sen Zha, Yan Zhang, Jiangyun Li

ECCV 2022 · Detection and segmentation › Open-set and few-shot detection · approach: Interactive and referring 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
decrease alpha from 0.001 to 0.0001 (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET / FLOPs per case: ET 77.00 -> 78.84; FLOPs 787.80 -> 1,966.20Mixed
increase alpha from 0.001 to 0.01 (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET / FLOPs per case: ET 77.00 -> 77.21; FLOPs 787.80 -> 756.83No clear effect
add softmax on FLOPs in choice metric (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET: 75.57 -> 77.00Improved
add softmax on Dice scores in choice metric (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET / FLOPs per case: ET 78.75 -> 77.00; FLOPs 1,551.485 -> 787.80Mixed
use modified ShuffleNetV2 instead of MobileNetV2 as Decision Network (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)overall FLOPs per case: 5,061.31 -> 787.80Improved
use modified ShuffleNetV2 instead of ShuffleNetV2 as Decision Network (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)overall FLOPs per case: 1,463.73 -> 787.80Improved
add candidate networks to skip-only bank (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET: 9.09 -> 75.60Improved
increase candidates from 2 to 4 (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET: 75.60 -> 77.00Improved
increase candidates from 4 to 6 (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET / FLOPs per case: ET 77.00 -> 78.71; FLOPs 787.80 -> 1,274.70Mixed
use joint training instead of individual training of candidates (ablation)Med-DANet · BraTS 2019 train 5-fold (ablation)Dice ET: 75.73 -> 78.75Improved
use dynamic selection instead of single largest candidate M4 (setting 1) (ablation)M4 · BraTS 2019 train 5-fold, setting 1 (slide window)FLOPs per case: 12,347.92 -> 1,551.485Improved
use dynamic selection instead of single candidate M4 (setting 2) (ablation)M4 · BraTS 2019 train 5-fold, setting 2 (full resolution)Dice ET: 76.48 -> 77.73Improved

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