VideoMem

该数据集名为VideoMem，包含10,000个无声的七秒钟视频，每个视频都附有短期和长期记忆性评分。尽管今年任务的评分中省略了长期评分，但仍有7,000个视频作为训练集发布，1,500个用于验证，另外1,500个则构成官方测试集。整个数据集规模达到10,000个视频，任务内容是预测视频的短期记忆性。

This dataset is named VideoMem, which comprises 10,000 silent 7-second videos, each paired with short-term and long-term memorability scores. Although long-term scores were omitted from this year's task annotations, 7,000 videos are released as the training set, 1,500 for validation, and another 1,500 form the official test set. The full dataset totals 10,000 videos, and the core task associated with this benchmark is to predict the short-term memorability of the videos.

搜集汇总

背景与挑战

背景概述

VideoMem数据集是一个包含10,000个无声视频片段及其记忆性评分的集合，用于视频记忆性理解和预测研究。数据集提供了短期和长期记忆性评分，并附带多种预提取的视频特征，如C3D和HoG，支持多媒体内容记忆性预测方法的评估。