让深度网络卷积核像传统滤波器一样归一化，提升鲁棒性。
Normalize Filters! Classical Wisdom for Deep Vision
- 引入滤波器归一化+可学习缩放偏移，模仿批归一化
- 在真实与人工光照变化下性能显著提升，超越CLIP
- 适合追求模型鲁棒性与泛化能力的研究者
传统图像滤波器（如均值或差分）经过精心归一化以保证一致性、可解释性，并避免亮度偏移、光晕或振铃等伪影。相比之下，深度网络端到端学习的卷积核缺乏此类约束。尽管它们可能类似小波或边缘/斑点检测器，但并未进行归一化。因此，当图像经历大气传输时，其响应会失真，导致错误结果。我们提出滤波器归一化，再结合可学习缩放和偏移，类似于批归一化。这一简单而有效的方法使滤波器具备大气等变性，实现共域对称性。将经典滤波原理融入深度学习（适用于卷积神经网络及依赖卷积的视觉变压器），该方法在人工与自然强度变化基准上取得显著提升。我们的ResNet34甚至大幅超越CLIP。分析表明，未归一化的滤波器会降低性能，而归一化则正则化学习过程，促进多样性，提升鲁棒性与泛化能力。
原文摘要 · Abstract (English)
Classical image filters, such as those for averaging or differencing, are carefully normalized to ensure consistency, interpretability, and to avoid artifacts like intensity shifts, halos, or ringing. In contrast, convolutional filters learned end-to-end in deep networks lack such constraints. Although they may resemble wavelets and blob/edge detectors, they are not normalized in the same or any way. Consequently, when images undergo atmospheric transfer, their responses become distorted, leading to incorrect outcomes. We address this limitation by proposing filter normalization, followed by learnable scaling and shifting, akin to batch normalization. This simple yet effective modification ensures that the filters are atmosphere-equivariant, enabling co-domain symmetry. By integrating classical filtering principles into deep learning (applicable to both convolutional neural networks and convolution-dependent vision transformers), our method achieves significant improvements on artificial and natural intensity variation benchmarks. Our ResNet34 could even outperform CLIP by a large margin. Our analysis reveals that unnormalized filters degrade performance, whereas filter normalization regularizes learning, promotes diversity, and improves robustness and generalization.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。