N-MNIST 和 N-Caltech101

N-MNIST和N-Caltech101是由新加坡神经技术研究所(SINAPSE)创建的用于神经形态视觉研究的数据集。N-MNIST由70000个样本组成，源自MNIST数据集，每个样本大小为28x28像素，适用于快速算法测试和原型设计。N-Caltech101包含101个对象类别，平均图像大小为245x302像素，提供了更具挑战性的识别任务。这两个数据集通过模拟生物视觉系统的运动感知机制，将静态图像转换为神经形态数据，旨在解决神经形态视觉传感器数据稀缺的问题，并促进与传统计算机视觉算法的直接比较。N-MNIST and N-Caltech101 are datasets developed by the Singapore Institute of Neurotechnology (SINAPSE) for neuromorphic vision research. N-MNIST comprises 70,000 samples derived from the MNIST dataset, with each sample sized at 28×28 pixels, making it suitable for rapid algorithm testing and prototyping. N-Caltech101 includes 101 object categories, with an average image size of 245×302 pixels, delivering more challenging recognition tasks. By simulating the motion perception mechanism of biological visual systems, these two datasets convert static images into neuromorphic data, with the goals of addressing the shortage of neuromorphic visual sensor data and facilitating direct comparisons with traditional computer vision algorithms.

搜集汇总

背景与挑战

背景概述

N-MNIST和N-Caltech101是由新加坡神经技术研究所创建的神经形态视觉数据集。N-MNIST包含70000个28x28像素样本，源自MNIST，适合快速算法测试；N-Caltech101包含101个类别，平均图像尺寸为245x302像素，提供更具挑战性的识别任务。它们通过模拟生物视觉系统将静态图像转换为神经形态数据，旨在解决数据稀缺问题并促进与传统计算机视觉算法的比较。