Unverified50% confidenceFactExact time

标准CNN具备平移等变性但不具备旋转等变性，工程师通常依赖数据增强随机旋转训练样本来识别不同角度物体

1

Sources

50%

Confidence

Long-term

Relevance

8/22/2026

First Seen

Sources

Spherical Fourier Transform: How to Build 3D Rotation-Equivariant AI Models

Related Claims

Unverified数据增强通过对训练样本施加随机变换（如旋转、翻转、缩放、裁剪、颜色抖动等）来人工扩大训练集规模，其理论基础是正则化72% similar Unverified球面等变网络在小数据集上可以提升数倍的学习效率，且对训练中未见过的旋转姿态也能稳健处理70% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/784718

API

curl https://kongchang.com/api/v1/knowledge/claims/784718

MCP

get_claim(id=784718)