CIFAR10-DVS

CIFAR10-DVS 是用于对象分类的事件流数据集。来自 CIFAR-10 数据集的 10,000 个基于帧的图像通过基于事件的传感器转换为 10,000 个事件流，其分辨率为 128×128 像素。该数据集具有 10 个不同类别的中等难度。采用基于帧图像的重复闭环平滑（RCLS）运动来实现转换。由于变换，它们在连续时间内产生丰富的局部强度变化，这些变化由基于事件的相机的每个像素进行量化。

CIFAR10-DVS is an event-stream dataset for object classification. A total of 10,000 frame-based images from the CIFAR-10 dataset are converted into 10,000 event streams via an event-based sensor, with a resolution of 128×128 pixels. This dataset features 10 distinct categories with moderate difficulty. The conversion is implemented using the repeated closed-loop smooth (RCLS) motion based on the frame-based images. Due to this transformation, the images generate rich local intensity variations over continuous time, which are quantified by each pixel of the event-based camera.

搜集汇总

背景与挑战

背景概述

CIFAR10-DVS是一个用于对象分类的事件流数据集，通过将CIFAR-10数据集的10,000张图像转换为128×128分辨率的事件流生成，涵盖10个类别。该数据集采用重复闭环平滑（RCLS）运动转换方法，模拟基于事件的相机感知，产生丰富的局部强度变化，适用于事件数据分析和分类任务。