five

Replication Data for: Dynamics of facial actions for assessing smile genuineness

收藏
NIAID Data Ecosystem2026-03-12 收录
下载链接:
https://doi.org/10.7910/DVN/X5QGLA
下载链接
链接失效反馈
官方服务:
资源简介:
The packages contain data extracted from the UvA-NEMO and BBC databases. The data encompass frame-wise smile intensities, smile phases (1-onset, 2-apex, 3-offset, 0-none), values of the 17 action units (AUs) and their dynamics (1st order and 2nd order Delta alongside the r-adjusted variants, extracted from Omega=9 and Omega=27 windows lengths). The features (AU-wise and cross-AU) are extracted for individual smile phases (onset, apex, offset), as well as for all these three phases ("whole_smile") and for the entire sequence ("whole_sequence"). Each file contains a single feature vector, stored as a line of comma-separated values. In addition to that, the features were extracted in sliding windows of the size of 13 and 25 frames (with a stride of 6 and 12 frames, respectively) - here, for each file, multiple feature vectors are stored, one per line. AU-wise features: amplitude (for 17 AUs, same sequence as defined in the "action_units" folder), average value (x17), maximum value (x17), average delta for Omega=9 and for Omega=27 (17 pairs), maximum delta (x17 pairs with interlacing values of Omega). Overall, 119 features per each feature vector. Cross-AU features: minimum value in a Delta-difference signal between every pair of AUs for Omega=9 and Omega=27 for 136 subsequent pairs of AUs (AU1-AU2, AU1-AU4, ..., AU1-AU45, AU2-AU4, ..., AU26-AU45). The data are stored with interlacing values of Omega (AU1-AU2 for Omega=9, AU1-AU2 for Omega=27, AU1-AU4 for Omega=9, etc.). Then the same goes for maximum value in the Delta-difference signal, and time difference between maximum and minimum r-adjusted Delta between two AUs. Overall, 1088 features.
创建时间:
2021-03-08
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作