2023 Video Similarity Dataset
收藏资源简介:
2023视频相似性数据集是由Meta AI和Czech Technical University in Prague共同创建,专注于视频复制检测与定位问题。该数据集包含约40,000个视频,旨在通过模拟真实场景中的‘针在干草堆’设置,评估视频复制检测和定位的准确性。数据集通过多种视频编辑和转换技术生成,以确保数据的真实性和挑战性。应用领域包括内容审核、版权保护和媒体识别,旨在解决视频内容重复和误导性使用的问题。
The 2023 Video Similarity Dataset was jointly developed by Meta AI and the Czech Technical University in Prague, focusing on the task of video copy detection and localization. This dataset contains approximately 40,000 videos, aiming to evaluate the accuracy of video copy detection and localization by simulating the "needle in a haystack" setting in real-world scenarios. The dataset is generated using a variety of video editing and transformation techniques to ensure its authenticity and challenging nature. Its application areas include content moderation, copyright protection and media recognition, aiming to solve the problems of video content duplication and misleading usage.

- 1The 2023 Video Similarity Dataset and ChallengeMeta AI · 2023年



