A 28nm 2D/3D Unified Sparse Convolution Accelerator with Block-Wise Neighbor Searcher for Large-Scaled Voxel-Based Point Cloud Network

📋 论文概要

该论文提出一种28nm工艺的2D/3D统一稀疏卷积加速器，针对大规模体素化点云处理，采用块级邻居搜索器来高效支持2D和3D稀疏卷积（包括子流形和非子流形），解决了自动驾驶等场景中稀疏卷积计算和访问的瓶颈问题。

💡 主要创新点

• 首次在同一芯片上统一支持2D和3D稀疏卷积，兼顾子流形和非子流形模式。
• 提出块级邻居搜索器，通过分块策略降低不规则稀疏卷积中的邻居查找开销。

🏷 关键词

📄 原文摘要

important role in many emerging applications such as autonomous driving, visual navigation and virtual reality. Recent research shows that adopting 3D voxel-based sparse convolution (SCONV) as a backbone can achieve better performance than a point-based network in large-scale outdoor scenarios [1]. Moreover, 2D SCONVs are still necessary for Bird’s-Eye-View (BEV) neck layers or fusion with image processing. Hardware acceleration is needed for multiple key operations, including 3D submanifold SCONV (S-SCONV), 3D non-submanifold SCONV (N-SCONV) and 2D SCONV. Recently, several processors have been developed for point-based networks [2] or SCONV [3-5]. However, for large-scale voxel-based sparse networks, three key challenges have not been fully addressed thereby limiting practical application, as shown

👥 作者与机构

Wenyu Sun1, Xiaoyu Feng1, Chen Tang1, Shupei Fan1, Yixiong Yang1,

Jinshan Yue2, Huazhong Yang1, Yongpan Liu1 Tsinghua University, Beijing, China Chinese Academy of Sciences, Beijing, China 1 2 3D processing plays an