AI Paper Atlas

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

Cross-Modal Learning for Domain Adaptation in 3D Semantic Segmentation

Maximilian Jaritz, Tuan-Hung Vu, Raoul de Charette, Émilie Wirbel, Patrick Pérez

TPAMI 2022 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Adaptation for semantic segmentation · 80 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 cross-modal loss (xMUDA) without pseudo-labelssource-only baseline · nuScenes-Lidarseg Day/Night UDAmIoU (2D+3D): 63.3 -> 67.4Improved
add cross-modal loss (xMUDA) without pseudo-labelssource-only baseline · VirtualKITTI/SemanticKITTI UDAmIoU (2D+3D): 42.2 -> 48.2Improved
add pseudo-labels on top of cross-modal learning (ablation)xMUDA · A2D2/SemanticKITTI UDAmIoU (2D+3D): 44.0 -> 47.5Improved
add pseudo-labels on top of cross-modal learning (ablation)xMUDA · VirtualKITTI/SemanticKITTI UDAmIoU (2D+3D): 48.2 -> 52.0Improved
add pseudo-labels on top of cross-modal learning (ablation)xMUDA · nuScenes-Lidarseg Day/Night UDAmIoU (2D+3D): 67.4 -> 64.4Hurt
use cross-modal loss instead of uni-modal PLPL baseline · VirtualKITTI/SemanticKITTI UDAmIoU (2D+3D): 35.6 -> 48.2Improved
apply cross-modal loss on source domain as well as target (ablation)xMUDA · nuScenes-Lidarseg USA/Singapore UDAmIoU (2D and 3D): +1.6 mIoU on 2D and +1.3 on 3DImproved
use dual-head (mimicking head) instead of single head, lambda_t=0.1 (ablation)xMUDA · nuScenes-Lidarseg USA/Singapore UDAmIoUImproved
add cross-modal loss to supervised trainingsupervised segmentation · Waymo OD KRKmIoU (2D+3D): 59.2 -> 61.1Improved
add cross-modal loss to supervised trainingsupervised segmentation · nuScenes-Lidarseg SingaporemIoU (2D+3D): 79.6 -> 79.8No clear effect
use xMUDA Fusion architecture instead of vanilla late fusion UDAfusion model · A2D2/SemanticKITTI UDAmIoU: 38.6 -> 42.6Improved
add cross-modal learning to semi-supervised adaptation with pseudo-labelsbaseline S+T_l · A2D2/SemanticKITTI SSDAmIoU (2D+3D): 66.2 -> 70.7Improved

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