AI Paper Atlas

Research map / Label-efficient and robust learning / Test-time adaptation

Learning Instance-Specific Adaptation for Cross-Domain Segmentation

Yuliang Zou, Zizhao Zhang, Chun-Liang Li, Han Zhang, Tomas Pfister, Jia-Bin Huang

ECCV 2022 · Label-efficient and robust learning › Test-time adaptation · approach: Test-time adaptation of vision-language models · 7 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
learn per-channel vector calibration strength instead of one scalar (ablation)InstCal · GTA5 to CityscapesmIoU: 39.8 -> 41.1Improved
learn separate calibration vectors for mean and variance (ablation)InstCal-U · GTA5 to CityscapesmIoU: 41.1 -> 41.5Improved
learn calibration strength instead of fixed m=0.1 (ablation)InstCal-U · GTA5 to CityscapesmIoU: 40.1 -> 41.5Improved
train calibration with DeepAugment instead of default weak augmentation (ablation)InstCal-U · GTA5 to CityscapesmIoU: 39.7 -> 41.5Improved
use RandAugment instead of DeepAugment for calibration training (ablation)InstCal-C · GTA5 to CityscapesmIoU: 42.2 -> 40.0Hurt
pre-train with DeepAugment but no InstCal (ablation)pre-trained DeepLabv2 · GTA5 to CityscapesmIoU: 35.7 -> 31.7Hurt
use conditional (MLP-weighted basis) calibration instead of unconditionalInstCal-U · GTA5 to CityscapesmIoU: 41.5 -> 42.2 (GTA5 avg 40.3 -> 41.1); Cityscapes-source avg 45.9 -> 45.0Mixed
use 8 calibration basis instead of 16 for conditional calibration (ablation)InstCal-C · GTA5 to CityscapesmIoU: 40.7 -> 42.2Improved
use per-instance statistics instead of batch statistics (batch size 16) (ablation)InstCal-U · GTA5 to CityscapesmIoU: 40.7 -> 41.5Improved
add test-time entropy minimizationInstCal-U · GTA5 to CityscapesmIoU: 41.5 -> 44.1Improved
add test-time entropy minimizationInstCal-C · GTA5 to CityscapesmIoU: 42.2 -> 44.2Improved
add InstCal-U to non-DG pre-trained modelDeepLabv3+ ResNet-50 · GTA5 to 3 real domains (avg)mIoU: 27.9 -> 37.1Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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