English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Open

Presents the pdf analysis as premium feature

PDF Analysis

Insights

Presents the summarisation as premium feature

Summarisation

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Detecting action units (AUs) on human faces is challenging because variousAUs make subtle facial appearance change over various regions at differentscales. Current works have attempted to recognize AUs by emphasizing importantregions. However, the incorporation of expert prior knowledge into regiondefinition remains under-exploited, and current AU detection approaches do notuse regional convolutional neural networks (R-CNN) with expert prior knowledgeto directly focus on AU-related regions adaptively. By incorporating expertprior knowledge, we propose a novel R-CNN based model named AU R-CNN. Theproposed solution offers two main contributions: (1) AU R-CNN directly observesdifferent facial regions, where various AUs are located. Specifically, wedefine an AU partition rule which encodes the expert prior knowledge into theregion definition and RoI-level label definition. This design producesconsiderably better detection performance than existing approaches. (2) Weintegrate various dynamic models (including convolutional long short-termmemory, two stream network, conditional random field, and temporal actionlocalization network) into AU R-CNN and then investigate and analyze the reasonbehind the performance of dynamic models. Experiment results demonstrate that\textit{only} static RGB image information and no optical flow-based AU R-CNNsurpasses the one fused with dynamic models. AU R-CNN is also superior totraditional CNNs that use the same backbone on varying image resolutions.State-of-the-art recognition performance of AU detection is achieved. Thecomplete network is end-to-end trainable. Experiments on BP4D and DISFAdatasets show the effectiveness of our approach. The implementation code isavailable online.

AU R-CNN: Encoding Expert Prior Knowledge into R-CNN for Action Unit Detection