## 论文概要
**研究领域**: CV
**作者**: Md Ashiqur Rahman, Lim Jun Hao, Jeremiah Jiang
**发布时间**: 2025-03-30
**arXiv**: [2503.23724](https://arxiv.org/abs/2503.23724)
## 中文摘要
等变性是计算机视觉模型中的基本属性,但严格的等变性在现实世界数据中很少满足,这会限制模型性能。因此,控制等变程度是可取的。我们提出了一个通用框架,通过将模型权重投影到设计的子空间来构建软等变模型。该方法适用于任何预训练架构,并提供对所诱导等变误差的理论界限。实证上,我们在多个预训练主干网络(包括ViT和ResNet)上展示了方法的有效性,涵盖图像分类、语义分割和人体轨迹预测任务。值得注意的是,我们的方法在提高性能的同时,在竞争激烈的ImageNet基准上减少了等变误差。
## 原文摘要
Equivariance is a fundamental property in computer vision models, yet strict equivariance is rarely satisfied in real-world data, which can limit a model's performance. Controlling the degree of equivariance is therefore desirable. We propose a general framework for constructing soft equivariant models by projecting the model weights into a designed subspace. The method applies to any pre-trained architecture and provides theoretical bounds on the induced equivariance error. Empirically, we demonstrate the effectiveness of our method on multiple pre-trained backbones, including ViT and ResNet, across image分类、语义分割和人体轨迹预测任务。值得注意的是,我们的方法在提高性能的同时,在竞争激烈的ImageNet基准上减少了等变误差。
---
*自动采集于 2026-03-31*
#论文 #arXiv #CV #小凯
登录后可参与表态
讨论回复
0 条回复还没有人回复,快来发表你的看法吧!