arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2405.04537cs.CVcs.AIcs.GR
收藏

一种用于 SO(3)-等变网络的直观多频率特征表示

An intuitive multi-frequency feature representation for SO(3)-equivariant networks

KAIST(韩国科学技术院)

Dongwon Son, Jaehyung Kim, Sanghyeon Son, Beomjoon Kim

AI总结:

针对 Vector Neuron 仅用三维特征、表达能力受限的问题,本文提出可映射三维点到高维空间并辨别多频率信息的 SO(3)-等变特征表示,使 VN 捕捉更多细节。

摘要:

形状重建等三维视觉算法的应用仍然受限,因为它们要求输入处于固定的标准旋转。最近,一种简单的等变网络 Vector Neuron(VN)被提出,可方便地与最先进的三维神经网络(NN)架构结合使用。然而,其性能受到限制,因为它被设计为仅使用三维特征,而这不足以捕捉三维数据中存在的细节。本文提出一种等变特征表示,用于将三维点映射到高维特征空间。我们的特征能够辨别三维数据中存在的多个频率,这是为三维视觉任务设计富有表达力的特征的关键。该表示可用作 VN 的输入,结果表明,借助我们的特征表示,VN 能捕捉更多细节,克服了其原始论文中提出的局限。

英文摘要:

The usage of 3D vision algorithms, such as shape reconstruction, remains limited because they require inputs to be at a fixed canonical rotation. Recently, a simple equivariant network, Vector Neuron (VN) has been proposed that can be easily used with the state-of-the-art 3D neural network (NN) architectures. However, its performance is limited because it is designed to use only three-dimensional features, which is insufficient to capture the details present in 3D data. In this paper, we introduce an equivariant feature representation for mapping a 3D point to a high-dimensional feature space. Our feature can discern multiple frequencies present in 3D data, which is the key to designing an expressive feature for 3D vision tasks. Our representation can be used as an input to VNs, and the results demonstrate that with our feature representation, VN captures more details, overcoming the limitation raised in its original paper.

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。

↑