Research map / Continual learning / Domain and prompt-based
Multi-Modal Continual Test-Time Adaptation for 3D Semantic Segmentation
Haozhi Cao, Yuecong Xu, Jianfei Yang, Pengyu Yin, Shenghai Yuan, Lihua Xie
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 changed | Where | Measured | Effect |
|---|---|---|---|
| adaptive centroid-weighted fusion of raw and augmentation-average predictions (iMPA) instead of augmentation-average only (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 5.8% degradation using pure augmentation-average | Improved |
| add centroid-based weighting in iMPA and xMPF versus simple average fusion (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: more than 2.5% | Improved |
| update class-wise centroids by enqueuing confident target features (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 6.1% decrease without update | Improved |
| add stochastic restoring of pseudo-source features in CMQs (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 8.4% gap without restore | Improved |
| increase number of 3D augmentations from 1 to 5 (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 33.2 -> 36.6 | Improved |
| increase number of 2D augmentations from 1 to 5 (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 35.0 -> 35.3 | No clear effect |
| lower restoring probability to 0.3 (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 36.6 -> 34.5 | Hurt |
| add confidence filtering tau_cf (0.0 -> 0.8) (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 34.8 -> 36.6 | Improved |
| add class-wise contrastive loss (lambda_cts 0.0 -> 1.0) (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 35.8 -> 36.6 | Improved |
| add model-based stochastic restoration to MMTTA | MMTTA · SemanticKITTI-to-Synthia | mIoU: 25.6 -> 28.7 | Improved |
| evaluate at end of each sequence (E-Test) instead of immediately (ablation) | CoMAC · SemanticKITTI-to-Synthia | mIoU: 6.1% improvement | Improved |
Closest papers by meaning
- MM-TTA: Multi-Modal Test-Time Adaptation for 3D Semantic Segmentation (CVPR 2022 · similarity 0.91)
- Each Test Image Deserves A Specific Prompt: Continual Test-Time Adaptation for 2D Medical Image Segmentation (CVPR 2024 · similarity 0.87)
- Towards Robust Multimodal Open-set Test-time Adaptation via Adaptive Entropy-aware Optimization (ICLR 2025 · similarity 0.87)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.87)
- ViDA: Homeostatic Visual Domain Adapter for Continual Test Time Adaptation (ICLR 2024 · similarity 0.87)
- ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains (NeurIPS 2025 · similarity 0.87)
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.
