Unverified50% confidenceFactExact time
Vision Transformer在大规模数据集上超越CNN，但在小数据集上表现不如CNN，因其缺乏图像特有的归纳偏置
1
Sources
50%
Confidence
Long-term
Relevance
8/21/2026
First Seen
Sources
Related Claims
UnverifiedVision Transformer leverages its self-attention mechanism to capture long-range feature correlations in images, demonstrating advantages in fine-grained classification tasks.68% similarUnverifiedCNN能从原始像素中自动提取层次化特征，Transformer能从原始文本序列中学习语义表示，大幅降低了对手工特征工程的依赖68% similarUnverifiedAfter superimposing tiny perturbations that are nearly invisible to the human eye onto an image correctly classified by a neural network, the model output will produce an error with extremely high confidence.65% similarUnverified图像对抗攻击中，人类视觉系统是低频感知主导对高频微小扰动不敏感，而CNN（卷积神经网络）对高频纹理特征高度敏感，这一感知鸿沟造成了人眼正常、AI认错的现象65% similarUnverified神经编码模型通常以在ImageNet等数据集上预训练的CNN或视觉Transformer作为特征提取骨干64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/784085API
curl https://kongchang.com/api/v1/knowledge/claims/784085MCP
get_claim(id=784085)