A method that derives a sampling distribution over modality-missing scenarios from latent-space distortions improves fine-tuning performance for multimodal semantic segmentation on remote sensing datasets compared to uniform dropout and LoRA adaptation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Latent Space Guided Scenario Sampling for Multimodal Segmentation Under Missing Modalities
A method that derives a sampling distribution over modality-missing scenarios from latent-space distortions improves fine-tuning performance for multimodal semantic segmentation on remote sensing datasets compared to uniform dropout and LoRA adaptation.