Research map / Detection and segmentation / CNN features
QuadMamba: Learning Quadtree-based Selective Scan for Visual State Space Model
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes QuadMamba, a visual state space model with a quadtree-based selective scan that groups image tokens by spatial locality to handle visual data for classification, detection and segmentation.
Problem
Mamba's sequential selective scan, designed for language, handles visual data poorly because of spatial locality and adjacency in images and large variations in scale and sparsity of visual tokens.
Approach
Replace the flat scan with a quadtree-based selective scan that partitions the image hierarchically so that spatially adjacent tokens are scanned together, with local-global adaptation for visual state space modeling.
Key results
- Model variants Lite, Small, Base and Large range from 1.4 to 7.91 GFLOPs on ImageNet-1k according to the appendix architecture table; accuracy numbers are not in the provided text.
Experimental setup
Datasets: ImageNet-1k
Benchmarks: ImageNet-1k
Metrics: top-1 accuracy GFLOPs
Baselines: ResT Vision Mamba VMamba Swin Transformer DeiT
Builds on: Mamba Vision Mamba ResT
Closest papers by meaning
- Spatial-Mamba: Effective Visual State Space Models via Structure-Aware State Fusion (ICLR 2025 · similarity 0.91)
- Multi-Scale VMamba: Hierarchy in Hierarchy Visual State Space Model (NeurIPS 2024 · similarity 0.90)
- EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba (AAAI 2025 · similarity 0.90)
- VMamba: Visual State Space Model (NeurIPS 2024 · similarity 0.90)
- GroupMamba: Efficient Group-Based Visual State Space Model (CVPR 2025 · similarity 0.90)
- Vision Mamba Mender (NeurIPS 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.