# DisentangleFormer: Spatial-Channel Decoupling for Multi-Channel Vision

arXiv: https://arxiv.org/abs/2512.04314 · HTML: https://mlsanity.com/paper/2512.04314/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 38 references that are in the map.
- First posted on arXiv 3 Dec 2025.

## Builds on (most cited first, 26 of 26)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Squeeze-and-Excitation Networks](https://mlsanity.com/paper/1709.01507/) (2017)
- [Going Deeper with Convolutions](https://mlsanity.com/paper/1409.4842/) (2014)
- [A ConvNet for the 2020s](https://mlsanity.com/paper/2201.03545/) (2022)
- [Rethinking Atrous Convolution for Semantic Image Segmentation](https://mlsanity.com/paper/1706.05587/) (2017)
- [Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions](https://mlsanity.com/paper/2102.12122/) (2021)
- [Learning deep representations by mutual information estimation and maximization](https://mlsanity.com/paper/1808.06670/) (2018)
- [Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations](https://mlsanity.com/paper/1811.12359/) (2018)
- [ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders](https://mlsanity.com/paper/2301.00808/) (2023)
- [ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks](https://mlsanity.com/paper/1910.03151/) (2019)
- [CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows](https://mlsanity.com/paper/2107.00652/) (2021)
- [MaxViT: Multi-Axis Vision Transformer](https://mlsanity.com/paper/2204.01697/) (2022)
- [LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference](https://mlsanity.com/paper/2104.01136/) (2021)
- [Selective Kernel Networks](https://mlsanity.com/paper/1903.06586/) (2019)
- [GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond](https://mlsanity.com/paper/1904.11492/) (2019)
- [DaViT: Dual Attention Vision Transformers](https://mlsanity.com/paper/2204.03645/) (2022)
- [BAM: Bottleneck Attention Module](https://mlsanity.com/paper/1807.06514/) (2018)
- [SpectralFormer: Rethinking Hyperspectral Image Classification with Transformers](https://mlsanity.com/paper/2107.02988/) (2021)
- [MobileNetV4 -- Universal Models for the Mobile Ecosystem](https://mlsanity.com/paper/2404.10518/) (2024)
- [reBEN: Refined BigEarthNet Dataset for Remote Sensing Image Analysis](https://mlsanity.com/paper/2407.03653/) (2024)
- [Cascaded Recurrent Neural Networks for Hyperspectral Image Classification](https://mlsanity.com/paper/1902.10858/) (2019)
- [Benchmarking and scaling of deep learning models for land cover image classification](https://mlsanity.com/paper/2111.09451/) (2021)
- [HyperspectralMAE: The Hyperspectral Imagery Classification Model using Fourier-Encoded Dual-Branch Masked Autoencoder](https://mlsanity.com/paper/2505.05710/) (2025)

## Abstract (arXiv)

Vision Transformers face a fundamental limitation: standard self-attention jointly processes spatial and channel dimensions, leading to entangled representations that prevent independent modeling of structural and semantic dependencies. This problem is especially pronounced in hyperspectral imaging, from satellite hyperspectral remote sensing to infrared pathology imaging, where channels capture distinct biophysical or biochemical cues. We propose DisentangleFormer, an architecture that achieves robust multi-channel vision representation through principled spatial-channel decoupling. Motivated by information-theoretic principles of decorrelated representation learning, our parallel design enables independent modeling of structural and semantic cues while minimizing redundancy between spatial and channel streams. Our design integrates three core components: (1) Parallel Disentanglement: Independently processes spatial-token and channel-token streams, enabling decorrelated feature learning across spatial and spectral dimensions, (2) Squeezed Token Enhancer: An adaptive calibration module that dynamically fuses spatial and channel streams, and (3) Multi-Scale FFN: complementing global attention with multi-scale local context to capture fine-grained structural and semantic dependencies. Extensive experiments on hyperspectral benchmarks demonstrate that DisentangleFormer achieves state-of-the-art performance, consistently outperforming existing models on Indian Pine, Pavia University, and Houston, the large-scale BigEarthNet remote sensing dataset, as well as an infrared pathology dataset. Moreover, it retains competitive accuracy on ImageNet while reducing computational cost by 17.8% in FLOPs. The code will be made publicly available upon acceptance.