Pith. sign in

REVIEW 17 cited by

LocalMamba: Visual State Space Model with Windowed Selective Scan

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2403.09338 v1 pith:BS7W4L7W submitted 2024-03-14 cs.CV cs.AI

Tao Huang , Xiaohuan Pei , Shan You , Fei Wang , Chen Qian , Chang Xu This is my paper · ORCID
classification cs.CVcs.AI
keywords scanlocalvisionacrosscapturingdependencieseffectivelylocalmamba
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Recent advancements in state space models, notably Mamba, have demonstrated significant progress in modeling long sequences for tasks like language understanding. Yet, their application in vision tasks has not markedly surpassed the performance of traditional Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs). This paper posits that the key to enhancing Vision Mamba (ViM) lies in optimizing scan directions for sequence modeling. Traditional ViM approaches, which flatten spatial tokens, overlook the preservation of local 2D dependencies, thereby elongating the distance between adjacent tokens. We introduce a novel local scanning strategy that divides images into distinct windows, effectively capturing local dependencies while maintaining a global perspective. Additionally, acknowledging the varying preferences for scan patterns across different network layers, we propose a dynamic method to independently search for the optimal scan choices for each layer, substantially improving performance. Extensive experiments across both plain and hierarchical models underscore our approach's superiority in effectively capturing image representations. For example, our model significantly outperforms Vim-Ti by 3.1% on ImageNet with the same 1.5G FLOPs. Code is available at: https://github.com/hunto/LocalMamba.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 17 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. UIS-Mamba: Exploring Mamba for Underwater Instance Segmentation via Dynamic Tree Scan and Hidden State Weaken

    cs.CV 2025-08 conditional novelty 6.0 of 10

    UIS-Mamba applies a Mamba state space backbone with dynamic tree scanning and background-hidden-state suppression to achieve SOTA underwater instance segmentation.

  2. AtrousMamaba: An Atrous-Window Scanning Visual State Space Model for Remote Sensing Change Detection

    cs.CV 2025-07 conditional novelty 6.0 of 10

    An atrous-window scanning strategy improves Mamba-based change detection on six remote sensing benchmarks, showing visual state space models can capture fine local details alongside global context.

  3. Training-free Token Reduction for Vision Mamba

    cs.CV 2025-07 conditional novelty 6.0 of 10

    MTR uses Mamba's timescale parameter Δ as a token importance score to merge unimportant tokens, giving training-free inference speedups with small accuracy loss.

  4. MUG: Pseudo Labeling Augmented Audio-Visual Mamba Network for Audio-Visual Video Parsing

    cs.CV 2025-07 conditional novelty 6.0 of 10

    MUG combines manually corrected pseudo-labels, cross-modal random track recombination, and a Mamba-Transformer network to reach new state-of-the-art F1 scores on the LLP audio-visual video parsing benchmark.

  5. ECP-Mamba: An Efficient Multi-scale Self-supervised Contrastive Learning Method with State Space Model for PolSAR Image Classification

    cs.CV 2025-06 conditional novelty 6.0 of 10

    ECP-Mamba, a Mamba-based network with a spiral scan and multi-scale self-distillation, reports state-of-the-art PolSAR image classification accuracy at label rates as low as 0.2%.

  6. ASM-UNet: Adaptive Scan Mamba Integrating Group Commonalities and Individual Variations for Fine-Grained Segmentation

    cs.CV 2025-08 conditional novelty 5.0 of 10

    A Mamba-based segmentation network whose scan order is guided by a per-image learned score, plus a new fine-grained biliary tract dataset.

  7. HydraMamba: Multi-Head State Space Model for Global Point Cloud Learning

    cs.CV 2025-07 conditional novelty 5.0 of 10

    A state space model based point cloud network with shuffled Hilbert serialization, a convolutional bidirectional S6 branch, and multi-head S6 achieves new top scores on ModelNet40, ShapeNet, S3DIS, and ScanObjectNN.

  8. Few-Shot Object Detection via Spatial-Channel State Space Model

    cs.CV 2025-07 conditional novelty 5.0 of 10

    A Mamba-based channel sequence model combined with spatial attention improves few-shot object detection on VOC and COCO.

  9. FMOcc: TPV-Driven Flow Matching for 3D Occupancy Prediction with Selective State Space Model

    cs.CV 2025-07 conditional novelty 5.0 of 10

    FMOcc uses flow matching with tri-perspective view and selective state space layers to improve 3D occupancy prediction from two camera frames.

  10. MARL-MambaContour: Unleashing Multi-Agent Deep Reinforcement Learning for Active Contour Optimization in Medical Image Segmentation

    cs.CV 2025-06 conditional novelty 5.0 of 10

    A multi-agent Soft Actor-Critic framework with adaptive entropy and a Mamba policy network iteratively moves contour points to segment organs, reporting higher Dice and boundary scores on five datasets.

  11. MambaHash: Visual State Space Deep Hashing Model for Large-Scale Image Retrieval

    cs.CV 2025-06 conditional novelty 5.0 of 10

    MambaHash uses a grouped, multi-directional Mamba backbone for deep supervised hashing and reports the highest mean average precision on CIFAR-10 and IMAGENET, with marginal gains on NUS-WIDE.

  12. SEMA: a Scalable and Efficient Mamba like Attention via Token Localization and Averaging

    cs.CV 2025-06 conditional novelty 5.0 of 10

    SEMA combines window attention with global token averaging, motivated by a dispersion theorem for generalized attention, and reports 0.2 to 0.7 percent top-1 accuracy gains over comparable vision Mamba and MILA models.

  13. CMamba: Learned Image Compression with State Space Models

    eess.IV 2025-02 conditional novelty 5.0 of 10

    A hybrid CNN and Mamba (state space model) image compression codec reports BD-Rate savings of 14.95% to 18.83% over VVC with fewer parameters, FLOPs, and lower decoding time than the prior best learned method.

  14. HybridTM: Combining Transformer and Mamba for 3D Semantic Segmentation

    cs.CV 2025-07 conditional novelty 4.0 of 10

    A new 3D semantic segmentation architecture that interleaves attention and Mamba operators within each layer achieves small but consistent gains on indoor and outdoor benchmarks.

  15. CWNet: Causal Wavelet Network for Low-Light Image Enhancement

    cs.CV 2025-07 conditional novelty 4.0 of 10

    CWNet mixes wavelet-based frequency enhancement, Mamba-style high-frequency scanning, and two semantic consistency losses to produce competitive low-light image enhancement with 1.23 million parameters.

  16. MambaNeXt-YOLO: A Hybrid State Space Model for Real-time Object Detection

    cs.CV 2025-06 conditional novelty 4.0 of 10

    MambaNeXt-YOLO, a hybrid CNN-Mamba detector, reports 66.6% mAP on PASCAL VOC from scratch and 27.8% on DOTA v1.5, with marginal gains over Mamba-YOLO and lower frame rates than several lightweight YOLO variants.

  17. A Survey on Mamba Architecture for Vision Applications

    cs.CV 2025-02 conditional novelty 1.0 of 10

    A survey of Mamba-based vision models that summarizes scanning mechanisms, key architectures, and benchmark results, contributing no new experimental findings.

Pith tools