HybridTM：融合 Transformer 与 Mamba 用于三维语义分割
HybridTM: Combining Transformer and Mamba for 3D Semantic Segmentation
AI总结：
针对 Transformer 复杂度高与 Mamba 三维特征表征不足的问题，本文提出 HybridTM 混合架构及内层混合策略，在多个室内外基准上实现三维语义分割 SOTA。
摘要：
基于 Transformer 的方法凭借强大的注意力机制，在三维语义分割中展现出卓越能力，但其二次复杂度限制了对大规模点云长程依赖关系的建模。近期基于 Mamba 的方法虽能以线性复杂度实现高效处理，却在提取三维特征时面临特征表征能力不足的问题。然而，如何有效融合二者的互补优势，仍是该领域尚未解决的挑战。本文提出 HybridTM，这是首个将 Transformer 与 Mamba 集成用于三维语义分割的混合架构。此外，我们提出内层混合策略（Inner Layer Hybrid Strategy），以更细粒度结合注意力与 Mamba，从而同时捕获长程依赖关系和细粒度局部特征。大量实验表明，HybridTM 在多种室内与室外数据集上均具备有效性和泛化能力。此外，HybridTM 在 ScanNet、ScanNet200 和 nuScenes 基准上取得了当前最优性能。代码将在 https://github.com/deepinact/HybridTM 公开。
英文摘要：
Transformer-based methods have demonstrated remarkable capabilities in 3D semantic segmentation through their powerful attention mechanisms, but the quadratic complexity limits their modeling of long-range dependencies in large-scale point clouds. While recent Mamba-based approaches offer efficient processing with linear complexity, they struggle with feature representation when extracting 3D features. However, effectively combining these complementary strengths remains an open challenge in this field. In this paper, we propose HybridTM, the first hybrid architecture that integrates Transformer and Mamba for 3D semantic segmentation. In addition, we propose the Inner Layer Hybrid Strategy, which combines attention and Mamba at a finer granularity, enabling simultaneous capture of long-range dependencies and fine-grained local features. Extensive experiments demonstrate the effectiveness and generalization of our HybridTM on diverse indoor and outdoor datasets. Furthermore, our HybridTM achieves state-of-the-art performance on ScanNet, ScanNet200, and nuScenes benchmarks. The code will be made available at https://github.com/deepinact/HybridTM.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
