# DIS2: Disentanglement Meets Distillation with Classwise Attention for Robust Remote Sensing Segmentation under Missing Modalities

arXiv: https://arxiv.org/abs/2601.13502 · HTML: https://mlsanity.com/paper/2601.13502/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Detection and dense prediction distillation, one of 12 directions in Knowledge distillation.
- Builds on 19 of its 32 references that are in the map.
- First posted on arXiv 20 Jan 2026.

## Placed in

- Knowledge distillation / [Detection and dense prediction distillation](https://mlsanity.com/distill/detection-and-dense-prediction-distillation/)

## Builds on (most cited first, 19 of 19)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Exploring Simple Siamese Representation Learning](https://mlsanity.com/paper/2011.10566/) (2020)
- [SMIL: Multimodal Learning with Severely Missing Modality](https://mlsanity.com/paper/2103.05677/) (2021)
- [ResViT: Residual vision transformers for multi-modal medical image synthesis](https://mlsanity.com/paper/2106.16031/) (2021)
- [Multi-modal Learning with Missing Modality via Shared-Specific Feature Modelling](https://mlsanity.com/paper/2307.14126/) (2023)
- [More Diverse Means Better: Multimodal Deep Learning Meets Remote Sensing Imagery Classification](https://mlsanity.com/paper/2008.05457/) (2020)
- [PMR: Prototypical Modal Rebalance for Multimodal Learning](https://mlsanity.com/paper/2211.07089/) (2022)
- [mmFormer: Multimodal Medical Transformer for Incomplete Multimodal Learning of Brain Tumor Segmentation](https://mlsanity.com/paper/2206.02425/) (2022)
- [Deep Multimodal Learning with Missing Modality: A Survey](https://mlsanity.com/paper/2409.07825/) (2024)
- [Hi-Net: Hybrid-fusion Network for Multi-modal MR Image Synthesis](https://mlsanity.com/paper/2002.05000/) (2020)
- [TransMed: Transformers Advance Multi-modal Medical Image Classification](https://mlsanity.com/paper/2103.05940/) (2021)
- [Multimodal Representation Learning by Alternating Unimodal Adaptation](https://mlsanity.com/paper/2311.10707/) (2023)
- [DrFuse: Learning Disentangled Representation for Clinical Multi-Modal Fusion with Missing Modality and Modal Inconsistency](https://mlsanity.com/paper/2403.06197/) (2024)
- [On the Limitations of Multimodal VAEs](https://mlsanity.com/paper/2110.04121/) (2021)
- [TIP: Tabular-Image Pre-training for Multimodal Classification with Incomplete Data](https://mlsanity.com/paper/2407.07582/) (2024)
- [Urban Land Cover Classification with Missing Data Modalities Using Deep Convolutional Neural Networks](https://mlsanity.com/paper/1709.07383/) (2017)
- [A Study of Dropout-Induced Modality Bias on Robustness to Missing Video Frames for Audio-Visual Speech Recognition](https://mlsanity.com/paper/2403.04245/) (2024)
- [General-Purpose Multimodal Transformer meets Remote Sensing Semantic Segmentation](https://mlsanity.com/paper/2307.03388/) (2023)
- [Filling the Gaps: A Multitask Hybrid Multiscale Generative Framework for Missing Modality in Remote Sensing Semantic Segmentation](https://mlsanity.com/paper/2509.11102/) (2025)

## Abstract (arXiv)

The efficacy of multimodal learning in remote sensing (RS) is severely undermined by missing modalities. The challenge is exacerbated by the RS highly heterogeneous data and huge scale variation. Consequently, paradigms proven effective in other domains often fail when confronted with these unique data characteristics. Conventional disentanglement learning, which relies on significant feature overlap between modalities (modality-invariant), is insufficient for this heterogeneity. Similarly, knowledge distillation becomes an ill-posed mimicry task where a student fails to focus on the necessary compensatory knowledge, leaving the semantic gap unaddressed. Our work is therefore built upon three pillars uniquely designed for RS: (1) principled missing information compensation, (2) class-specific modality contribution, and (3) multi-resolution feature importance. We propose a novel method DIS2, a new paradigm shifting from modality-shared feature dependence and untargeted imitation to active, guided missing features compensation. Its core novelty lies in a reformulated synergy between disentanglement learning and knowledge distillation, termed DLKD. Compensatory features are explicitly captured which, when fused with the features of the available modality, approximate the ideal fused representation of the full-modality case. To address the class-specific challenge, our Classwise Feature Learning Module (CFLM) adaptively learn discriminative evidence for each target depending on signal availability. Both DLKD and CFLM are supported by a hierarchical hybrid fusion (HF) structure using features across resolutions to strengthen prediction. Extensive experiments validate that our proposed approach significantly outperforms state-of-the-art methods across benchmarks.