Rotation Equivariant Mamba for Vision Tasks
cs.CV
Submitted: 2026-03-10
Updated: 2026-10-29
Code: https://github.com/zhongchenzhao/EQ-VMamba
License: http://creativecommons.org/licenses/by/4.0/
The gist: Rotation equivariance is an important structural prior for a broad range of vision tasks, particularly in domains such as remote sensing and medical imaging where objects have no canonical…
Terminology
Abstract
Rotation equivariance is an important structural prior for a broad range of vision tasks, particularly in domains such as remote sensing and medical imaging where objects have no canonical orientation, yet it remains notably absent from current Mamba-based vision architectures. Despite the success of Mamba in natural language processing and its growing adoption in computer vision, existing visual Mamba models fail to account for rotational symmetry in their design. This omission renders them sensitive to image rotations, thereby constraining their robustness and cross-task generalization. To address this limitation, we incorporate rotation symmetry, an important geometric prior in many visual domains, into Mamba-based architectures. Specifically, we introduce EQ-VMamba, the first rotation equivariant visual Mamba architecture for vision tasks. The core components of EQ-VMamba include a carefully designed rotation equivariant cross-scan strategy and group Mamba blocks. Moreover, we provide a rigorous theoretical analysis of the intrinsic equivariance error, demonstrating that the proposed architecture enforces end-to-end rotation equivariance under the p4 rotation group. Extensive experiments across multiple benchmarks---including high-level image classification, mid-level semantic segmentation, and low-level image super-resolution---demonstrate that EQ-VMamba consistently improves rotation robustness and achieves superior or competitive performance compared to non-equivariant baselines on most benchmarks, while requiring 40-70% fewer parameters. These results indicate that embedding rotation equivariance not only effectively bolsters the robustness of visual Mamba models against rotation transformations, but also improves accuracy on data exhibiting rotational symmetry, with substantially improved parameter efficiency. Code is available at https://github.com/zhongchenzhao/EQ-VMamba.
Sources
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Efficiently Modeling Long Sequences with Structured State Spaces
- Simplified State Space Layers for Sequence Modeling
- Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model
- Polyline Path Masked Attention for Vision Transformer
- Steerable Transformers for Volumetric Data
- Steerable CNNs
- Group Equivariant Stand-Alone Self-Attention For Vision
- Generating Long Sequences with Sparse Transformers
- RWKV: Reinventing RNNs for the Transformer Era
- Retentive Network: A Successor to Transformer for Large Language Models
- MambaIRv2: Attentive State Space Restoration
- DiM: Diffusion Mamba for Efficient High-Resolution Image Synthesis
- Learning Invariant Representations with Local Transformations
- HexaConv
- Decoupled Weight Decay Regularization
- LoveDA: A Remote Sensing Land-Cover Dataset for Domain Adaptive Semantic Segmentation
- On Efficient Transformer-Based Image Pre-training for Low-Level Vision
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs