AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach

Xiao Song, Guorun Yang, Xinge Zhu, Hui Zhou, Yuexin Ma, Zhe Wang, Jianping Shi

IJCV 2022 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Geometric alignment and synthetic-to-real adaptation · 16 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 progressive color transfer (ablation)Ada-PSMNet · KITTI 2015 trainD1-error (%): 12.1 -> 5.4Improved
add cost normalization (ablation)Ada-PSMNet · KITTI 2015 trainD1-error (%): 12.1 -> 9.1Improved
add self-supervised occlusion-aware reconstruction (ablation)Ada-PSMNet · KITTI 2015 trainD1-error (%): 4.7 -> 3.5Improved
use progressive update (gamma 0.95) instead of none (gamma 1) (ablation)progressive color transfer · KITTI trainD1-error (%): 6.2 -> 5.4Improved
combine channel and pixel normalization (ablation)cost normalization · KITTI trainD1-error (%): 12.1 -> 9.1Improved
replace general normalization layers with cost norm (ablation)PSMNet · KITTI trainD1-error (%): 10.7 -> 9.1Improved
use progressive color transfer instead of Reinhard color transfer (ablation)PSMNet · KITTI trainD1-error (%): 6.2 -> 5.4Improved
use progressive color transfer instead of GAN-based transfer (ablation)PSMNet · KITTI trainD1-error (%): 8.0 -> 5.4Improved
add high-level auxiliary tasks (rotation, flip, patch localization) (ablation)PSMNet baseline · KITTI trainD1-error (%): 12.1 -> 14.6Hurt
add source-domain occlusion mask loss (ablation)Ada-PSMNet · DrivingStereo trainD1-error (%): 5.4 -> 5.1Improved
add disparity smoothness loss (ablation)Ada-PSMNet · DrivingStereo trainD1-error (%): 5.2 -> 5.1Improved
use fewer target-domain training pairs (1/5) (ablation)Ada-PSMNet · DrivingStereo trainD1-error (%): 5.4 -> 5.1No clear effect

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