D-PAttNet: Dynamic Patch-Attentive Deep Network for Action Unit Detection

Abstract

Facial action units (AUs) relate to specific local facial regions. Recent efforts in automated AU detection have focused on learning the facial patch representations to detect specific AUs. These efforts have encountered three hurdles. First, they implicitly assume that facial patches are robust to head rotation; yet non-frontal rotation is common. Second, mappings between AUs and patches are defined a priori, which ignores co-occurrences among AUs. And third, the dynamics of AUs are either ignored or modeled sequentially rather than simultaneously as in human perception. Inspired by recent advances in human perception, we propose a dynamic patch-attentive deep network, called D-PAttNet, for AU detection that (i) controls for 3D head and face rotation, (ii) learns mappings of patches to AUs, and (iii) models spatiotemporal dynamics. D-PAttNet approach significantly improves upon existing state of the art.

Online biology degree program broadens access for women, first-generation to college, and low-income students, but grade disparities remain

“…The cause of the overall lower online grades is difficult to identify definitively within this study. Our finding is consistent with some prior work [ 11 , 12 , 33 , 34 ], but not all [ 14 , 32 ]. There are several factors that could be influencing the lower online grades in our study.…”

“…Finally, although our models do account for the students' college performance, they do not fully control for academic preparation, thus the online effect could be partially explained by lesser college preparedness. This would be consistent with Cavanaugh & Jacquemin (2015) who, while finding a negligible difference in online vs. face-to-face courses, also found that online courses tend to magnify the effect of differences in GPA, with low GPA students performing even worse and high GPA students performing even better [14,32].…”

“…Another factor to consider is student performance in online courses. While several studies report comparable learning outcomes for students in online courses compared to face-to-face courses [ 14 , 32 ], other studies show worse student performance in online courses [ 11 , 12 , 33 , 34 ]. Xu & Xu [ 18 ] suggest that these differences in results might be explained by the selectivity of institutions studied.…”

Action Unit Driven Facial Expression Synthesis from a Single Image with Patch Attentive GAN

“…• We propose a novel discriminator architecture that can improve both expressiveness and photo-realism for AU synthesis. Specifically, we first modify the state-of-the-art patch-based AU occurrence detection model PAttNet from [EJC19] for AU intensity estimation and treat it as the core critic in our GAN framework. Then WGAN-GP is replaced by least square GAN [MLX*17] (LSGAN) to avoid the removing of batch normalization layers.…”

Untitled

“…Another factor to consider is student performance in online courses. While several studies report comparable learning outcomes for students in online courses compared to face-toface courses [14,32], other studies show worse student performance in online courses [11,12,33,34]. Xu & Xu [18] suggest that these differences in results might be explained by the selectivity of institutions studied.…”