DAMamba: Vision State Space Model with Dynamic Adaptive Scan
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917927436943360 |
|---|---|
| author | Li, Tanzhe Li, Caoshuo Lyu, Jiayi Pei, Hongjuan Zhang, Baochang Jin, Taisong Ji, Rongrong |
| author_facet | Li, Tanzhe Li, Caoshuo Lyu, Jiayi Pei, Hongjuan Zhang, Baochang Jin, Taisong Ji, Rongrong |
| contents | State space models (SSMs) have recently garnered significant attention in computer vision. However, due to the unique characteristics of image data, adapting SSMs from natural language processing to computer vision has not outperformed the state-of-the-art convolutional neural networks (CNNs) and Vision Transformers (ViTs). Existing vision SSMs primarily leverage manually designed scans to flatten image patches into sequences locally or globally. This approach disrupts the original semantic spatial adjacency of the image and lacks flexibility, making it difficult to capture complex image structures. To address this limitation, we propose Dynamic Adaptive Scan (DAS), a data-driven method that adaptively allocates scanning orders and regions. This enables more flexible modeling capabilities while maintaining linear computational complexity and global modeling capacity. Based on DAS, we further propose the vision backbone DAMamba, which significantly outperforms current state-of-the-art vision Mamba models in vision tasks such as image classification, object detection, instance segmentation, and semantic segmentation. Notably, it surpasses some of the latest state-of-the-art CNNs and ViTs. Code will be available at https://github.com/ltzovo/DAMamba. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_12627 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DAMamba: Vision State Space Model with Dynamic Adaptive Scan Li, Tanzhe Li, Caoshuo Lyu, Jiayi Pei, Hongjuan Zhang, Baochang Jin, Taisong Ji, Rongrong Computer Vision and Pattern Recognition State space models (SSMs) have recently garnered significant attention in computer vision. However, due to the unique characteristics of image data, adapting SSMs from natural language processing to computer vision has not outperformed the state-of-the-art convolutional neural networks (CNNs) and Vision Transformers (ViTs). Existing vision SSMs primarily leverage manually designed scans to flatten image patches into sequences locally or globally. This approach disrupts the original semantic spatial adjacency of the image and lacks flexibility, making it difficult to capture complex image structures. To address this limitation, we propose Dynamic Adaptive Scan (DAS), a data-driven method that adaptively allocates scanning orders and regions. This enables more flexible modeling capabilities while maintaining linear computational complexity and global modeling capacity. Based on DAS, we further propose the vision backbone DAMamba, which significantly outperforms current state-of-the-art vision Mamba models in vision tasks such as image classification, object detection, instance segmentation, and semantic segmentation. Notably, it surpasses some of the latest state-of-the-art CNNs and ViTs. Code will be available at https://github.com/ltzovo/DAMamba. |
| title | DAMamba: Vision State Space Model with Dynamic Adaptive Scan |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2502.12627 |