论文标题
培训具有时空片段的神经形态数据上强大的尖峰神经网络
Training Robust Spiking Neural Networks on Neuromorphic Data with Spatiotemporal Fragments
论文作者
论文摘要
神经形态视觉传感器（事件摄像机）固有地适合尖峰神经网络（SNN），并为此仿生模型提供新的神经形态视觉数据。由于时空特征，需要新的数据增强来处理这些相机的非常规视觉信号。在本文中，我们提出了一种新型事件时空片段（ESTF）增强方法。它通过漂移或倒置时空事件流的片段来保留神经形态数据的连续性，以模拟亮度变化的干扰，从而导致更强大的尖峰神经网络。对盛行的神经形态数据集进行了广泛的实验。事实证明，ESTF对纯几何变换提供了实质性改进，并优于其他事件数据增强方法。值得注意的是，具有ESTF的SNN在CIFAR10-DVS数据集上实现了83.9％的最新精度。
Neuromorphic vision sensors (event cameras) are inherently suitable for spiking neural networks (SNNs) and provide novel neuromorphic vision data for this biomimetic model. Due to the spatiotemporal characteristics, novel data augmentations are required to process the unconventional visual signals of these cameras. In this paper, we propose a novel Event SpatioTemporal Fragments (ESTF) augmentation method. It preserves the continuity of neuromorphic data by drifting or inverting fragments of the spatiotemporal event stream to simulate the disturbance of brightness variations, leading to more robust spiking neural networks. Extensive experiments are performed on prevailing neuromorphic datasets. It turns out that ESTF provides substantial improvements over pure geometric transformations and outperforms other event data augmentation methods. It is worth noting that the SNNs with ESTF achieve the state-of-the-art accuracy of 83.9\% on the CIFAR10-DVS dataset.
