通过可控噪声注入提升神经网络对事件数据的抗噪能力
Learning from Noise: Enhancing DNNs for Event-Based Vision through Controlled Noise Injection
- 在训练时主动注入噪声，让模型学会识别真实信号
- 在多个数据集上平均准确率最高，优于传统滤波方法
- 适用于各类神经网络，特别适合高速或弱光场景
事件传感器相比传统帧式相机在快速运动或复杂光照下具有显著优势，但事件数据常含大量噪声，影响深度学习模型性能。传统方法依赖滤波算法清理噪声，可能误删有效信息。本文提出一种新型噪声注入训练方法，在训练阶段引入可控噪声，使神经网络学习抗噪表征。我们在N-Caltech101、N-Cars和Mini N-ImageNet等多个基准数据集上，使用CNN、ViT、SNN和GCN等多种架构进行评估。结果表明，该方法在不同噪声强度下均保持稳定性能，显著优于事件滤波技术，并达到最高平均分类准确率，为事件视觉分类系统提供了一种可靠的替代滤波方案。
原文摘要 · Abstract (English)
Event-based sensors offer significant advantages over traditional frame-based cameras, especially in scenarios involving rapid motion or challenging lighting conditions. However, event data frequently suffers from considerable noise, negatively impacting the performance and robustness of deep learning models. Traditionally, this problem has been addressed by applying filtering algorithms to the event stream, but this may also remove some of relevant data. In this paper, we propose a novel noise-injection training methodology designed to enhance the neural networks robustness against varying levels of event noise. Our approach introduces controlled noise directly into the training data, enabling models to learn noise-resilient representations. We have conducted extensive evaluations of the proposed method using multiple benchmark datasets (N-Caltech101, N-Cars, and Mini N-ImageNet) and various network architectures, including Convolutional Neural Networks, Vision Transformers, Spiking Neural Networks, and Graph Convolutional Networks. Experimental results show that our noise-injection training strategy achieves stable performance over a range of noise intensities, consistently outperforms event-filtering techniques, and achieves the highest average classification accuracy, making it a viable alternative to traditional event-data filtering methods in an object classification system. Code: https://github.com/vision-agh/DVS_Filtering
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。