Micro-video Thumbnail Selection Dataset

Micro-video Thumbnail Selection Dataset是由Bo Liu构建的一个大规模数据集，旨在通过分析用户兴趣来优化微视频缩略图的选择。该数据集从Instagram平台收集了35,148个微视频及其缩略图，通过分析用户评论中的高频词汇来捕捉用户兴趣。数据集的创建过程包括收集高点击量的微视频和提取候选帧，以及通过视觉质量和代表性筛选帧。该数据集主要应用于微视频平台的用户体验优化，通过提高缩略图的吸引力来增加视频的点击率。The Micro-video Thumbnail Selection Dataset is a large-scale dataset constructed by Bo Liu, which aims to optimize micro-video thumbnail selection by analyzing user interests. This dataset collects 35,148 micro-videos and their corresponding thumbnails from the Instagram platform, and captures user interests by analyzing high-frequency vocabulary in user comments. The dataset construction process includes collecting high-click-through micro-videos, extracting candidate frames, and screening frames based on visual quality and representativeness. This dataset is mainly applied to optimize user experience on micro-video platforms, increasing video click-through rates by improving the attractiveness of thumbnails.

搜集汇总

背景与挑战

背景概述

该数据集是一个大规模微视频缩略图选择数据集，由Bo Liu构建，从Instagram平台收集了35,148个微视频及其缩略图，通过分析用户评论高频词汇来捕捉用户兴趣，以优化缩略图选择。它主要用于微视频平台的用户体验优化，通过提高缩略图吸引力来增加视频点击率，数据收集过程包括高点击量视频筛选和基于视觉质量与代表性的帧提取。