YUP++ (YUP++ Dynamic Scenes dataset)

本网页提供动态场景视频数据集。这个新数据集从 20 个场景类别中抽取样本，同时涵盖范围广泛的条件，包括由自然场景类别差异、季节和昼夜变化以及观看参数引起的条件。每个类的缩略图示例如下所示。对于数据集中的每个场景类，有 60 个彩色视频，给定类没有两个样本取自同一物理场景。每个类别中的一半视频是使用静态摄像机获取的，一半是使用移动摄像机获取的，摄像机运动包括平移、倾斜、缩放和抖动。每个类都有静态和移动相机实例，可以系统地考虑这个变量在场景分类中所起的作用。所有视频均已使用 ffmpeg 视频库使用 H.264 编解码器进行压缩。每个视频的持续时间为 5 秒，原始帧速率介于每秒 24 到 30 帧之间。所有这些都已调整为最大宽度为 480 像素，同时保留其原始纵横比。

This webpage provides a dynamic scene video dataset. This novel dataset draws samples from 20 scene categories and encompasses a broad spectrum of conditions, including those arising from variations between natural scene categories, seasonal and diurnal changes, as well as viewing parameters. Thumbnail examples for each category are presented below. For every scene class in this dataset, there are 60 color videos, and no two samples from a single class are sourced from the same physical scene. Half of the videos in each category are captured using static cameras, while the remaining half are captured with moving cameras, with camera movements encompassing panning, tilting, zooming, and shaking. Each class features both static and moving camera setups, enabling the systematic analysis of the role this variable plays in scene classification. All videos are compressed using the H.264 codec via the FFmpeg video library. Each video has a duration of 5 seconds, with the original frame rate ranging from 24 to 30 frames per second. All videos have been resized to a maximum width of 480 pixels while maintaining their original aspect ratios.

搜集汇总

背景与挑战

背景概述

YUP++是一个动态场景视频数据集，包含20个场景类别，每个类别有60个视频，涵盖静态和移动摄像机拍摄的场景。视频经过压缩和分辨率调整，适用于动态场景识别研究。