HuggingFace Daily Papers·· 2 天前
QUAN：用四元数近似网络提升图像分类与旋转目标检测
Quaternion Approximate Networks for Enhanced Image Classification and Oriented Object Detection
AI 导读
论文提出四元数近似网络 QUAN，通过 Hamilton 积分解用实数运算近似四元数卷积，并配合自定义 CUDA kernel 实现，同时引入独立四元数批归一化（IQBN）稳定训练。在 CIFAR-10/100、ImageNet 分类和 COCO、DOTA 检测任务上，QUAN 以更少参数取得更高精度和更快收敛，并在四元数 CNN 的检测任务上达到 SOTA。
正文 · 原文
Abstract:This paper introduces Quaternion Approximate Networks (QUAN), a novel deep learning framework that leverages quaternion algebra for rotation equivariant image classification and object detection. Unlike conventional quaternion neural networks attempting to operate entirely in the quaternion domain, QUAN approximates quaternion convolution through Hamilton product decomposition using real-valued operations. This approach preserves geometric properties while enabling efficient implementation with custom CUDA kernels. We introduce Independent Quaternion Batch Normalization (IQBN) for training stability and extend quaternion operations to spatial attention mechanisms. QUAN is evaluated on image classification (CIFAR-10/100, ImageNet), object detection (COCO, DOTA), and robotic perception tasks. In classification tasks, QUAN achieves higher accuracy with fewer parameters and faster convergence compared to existing convolution and quaternion-based models. For objection detection, QUAN demonstrates improved parameter efficiency and rotation handling over standard Convolutional Neural Networks (CNNs) while establishing the SOTA for quaternion CNNs in this downstream task. These results highlight its potential for deployment in resource-constrained robotic systems requiring rotation-aware perception and application in other domains. Code: this https URL QUANpaper/
| Comments: | IROS 2025 Oral, 8 pages, 5 figures |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO) |
| Cite as: | arXiv:2509.05512 [cs.CV] |
| (or arXiv:2509.05512v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2509.05512 arXiv-issued DOI via DataCite |
|
| Journal reference: | 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Hangzhou, China, 2025, pp. 4240-4247 |
| Related DOI: | https://doi.org/10.1109/IROS60139.2025.11247562
DOI(s) linking to related resources |
Submission history
From: Bryce Grant [view email]
[v1]
Fri, 5 Sep 2025 21:41:40 UTC (7,063 KB)
[v2]
Tue, 27 Oct 2026 17:53:05 UTC (7,060 KB)
来源：HuggingFace Daily Papers · arxiv.org