AI Paper Atlas

Research map / Detection and segmentation / CNN features

Distribution-Aware Margin Calibration for Semantic Segmentation in Images

Litao Yu, Zhibin Li, Min Gang Xu, Yongsheng Gao, Jiebo Luo, Jian Zhang

IJCV 2021 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 20 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
replace cross-entropy with margin calibrationCOPLE-Net · Robotic Instrument testmIoU: 66.2 -> 72.5Improved
replace focal loss with margin calibrationCOPLE-Net · Robotic Instrument testmIoU: 69.5 -> 72.5Improved
use Dice loss as sole objectiveCOPLE-Net · Robotic Instrument testmIoU: 66.2 -> 27.3Hurt
use Tversky loss as sole objectiveCOPLE-Net · Robotic Instrument testmIoU: 66.2 -> 43.8Hurt
replace cross-entropy with margin calibrationDeepLab v3+ · COCO-Stuff 10K valmIoU: 34.1 -> 35.5Improved
replace Lovasz-softmax with margin calibrationDeepLab v3+ · PASCAL VOC2012 valmIoU: 78.5 -> 78.6No clear effect
fine-tune with margin calibration instead of cross-entropyDeepLab v3+ · SceneParse150 val (flip+multi-scale)mIoU: 39.1 -> 40.2Improved
fine-tune with margin calibration instead of cross-entropyDeepLab v3+ · Cityscapes val (flip+multi-scale)mIoU: 79.4 -> 81.1Improved
fine-tune with margin calibration instead of cross-entropyDeepLab v3+ · BDD100K val (flip+multi-scale)mIoU: 65.7 -> 66.1Improved
fine-tune with margin calibration instead of cross-entropyDeepLab v3+ · Mapillary Vistas val (flip+multi-scale)mIoU: 49.8 -> 51.1Improved
add flipping and multi-scale inference (ablation)DeepLab v3+ margin calibration · SceneParse150 valmIoU: 38.5 -> 40.2Improved
replace cross-entropy with margin calibration (no coarse pretrain)DeepLab v3+ SEResNeXt-50 · Cityscapes testIoU class: 80.5 -> 81.5Improved

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