CIFAR10-DVS 物体分类事件流数据集

CIFAR 全称 Canadian Institutes for Advanced Research，DVS 全称 Dynamic Vision Sensor，CIFAR10-DVS 是关于物体分类的事件流数据集。该数据集有 10 个类别（飞机、汽车、鸟、猫、鹿、狗、青蛙、马、船、卡车），共 1 万个事件流，属于中等难度。这些事件流由 CIFAR-10 数据集中基于帧的图像通过基于事件的图像传感器转换而得，采用图像重复闭环平滑（RCLS）运动来实现转换。由于转换，图像在连续时间内产生丰富的局部强度变化，这些变化被事件相机的以像素为基准量化。

CIFAR is the abbreviation for Canadian Institutes for Advanced Research, and DVS stands for Dynamic Vision Sensor. CIFAR10-DVS is an object classification event-stream dataset. It includes 10 categories (airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck) and a total of 10,000 event streams, and it is a moderately difficult dataset. These event streams are generated by converting frame-based images from the original CIFAR-10 dataset via event-based image sensors, with the repeated closed-loop smooth (RCLS) motion technique adopted for the conversion. As a result of this conversion, the images generate rich local intensity variations across continuous time, which are quantified on a per-pixel basis by the event camera.

搜集汇总

背景与挑战

背景概述

CIFAR10-DVS是一个用于物体分类的事件流数据集，包含10个类别共1万个事件流，由CIFAR-10图像通过基于事件相机的转换方法生成。该数据集属于中等难度，适用于图像分类任务。