VIDI
收藏资源简介:
VIDI数据集是由伊斯坦布尔技术大学创建的一个包含4534个视频片段的大型视频数据集,涵盖43种不同的事故类别,如地震、洪水、交通事故等。每个事故类别平均包含约100个视频,视频平均时长为十秒。数据集的视频主要从YouTube收集,通过使用多种语言搜索以增加多样性。VIDI数据集的创建旨在利用视频中的时间信息提高事故分类的准确性,特别适用于研究自然灾害和事故的自动检测。通过提供丰富的视频数据,VIDI数据集支持研究者开发和评估利用时间信息的检测和分类系统。
The VIDI dataset is a large-scale video dataset created by Istanbul Technical University, containing 4,534 video clips across 43 distinct accident categories including earthquakes, floods, traffic accidents and others. Each category holds approximately 100 videos on average, with an average video duration of 10 seconds. The dataset's videos are primarily sourced from YouTube, with multilingual search strategies employed to boost diversity. The VIDI dataset was developed to leverage temporal information within videos to enhance the accuracy of accident classification, and is particularly suited for research on automated detection of natural disasters and accidents. By offering abundant video data, the VIDI dataset enables researchers to develop and evaluate detection and classification systems that utilize temporal information.

- 1VIDI: A Video Dataset of Incidents伊斯坦布尔技术大学 · 2022年



