SMIC-E-Long
收藏资源简介:
SMIC-E-Long数据集是由奥卢大学机器视觉与信号分析中心创建的一个新的挑战性基准,用于微表情检测。该数据集包含162个长视频,总计约350000帧,其中132个视频包含微表情样本,共有167个微表情样本和16个主题。视频分辨率为640×480,帧率为100fps。数据集的创建过程中,通过在每个微表情样本前后添加2000至3000个自然帧(约20秒)来扩展视频长度,从而形成约22秒的长视频。此外,还选择了一些不包含微表情但包含常规面部表情的视频片段,以模拟实际的微表情检测情况。SMIC-E-Long数据集旨在解决现有微表情数据集在视频长度和复杂面部行为方面的不足,为开发稳健的微表情检测算法提供更真实的环境。
The SMIC-E-Long dataset is a novel challenging benchmark for micro-expression detection, created by the Center for Machine Vision and Signal Analysis of the University of Oulu. It contains 162 long videos with a total of approximately 350,000 frames, among which 132 videos have micro-expression samples, totaling 167 micro-expression instances and involving 16 subjects. The videos have a resolution of 640×480 and a frame rate of 100 fps. During the dataset construction process, the video length was extended to around 22 seconds by adding 2000 to 3000 natural frames (about 20 seconds) before and after each micro-expression sample. In addition, some video clips that do not contain micro-expressions but include regular facial expressions were selected to simulate real-world micro-expression detection scenarios. The SMIC-E-Long dataset aims to address the limitations of existing micro-expression datasets in terms of video length and complex facial behaviors, providing a more realistic environment for developing robust micro-expression detection algorithms.




