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/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

While Convolutional Neural Networks (CNNs) have significantly boosted theperformance of face related algorithms, maintaining accuracy and efficiencysimultaneously in practical use remains challenging. The state-of-the-artmethods employ deeper networks for better performance, which makes it lesspractical for mobile applications because of more parameters and highercomputational complexity. Therefore, we propose an efficient multitask neuralnetwork, Alignment & Tracking & Pose Network (ATPN) for face alignment, facetracking and head pose estimation. Specifically, to achieve better performancewith fewer layers for face alignment, we introduce a shortcut connectionbetween shallow-layer and deep-layer features. We find the shallow-layerfeatures are highly correspond to facial boundaries that can provide thestructural information of face and it is crucial for face alignment. Moreover,we generate a cheap heatmap based on the face alignment result and fuse it withfeatures to improve the performance of the other two tasks. Based on theheatmap, the network can utilize both geometric information of landmarks andappearance information for head pose estimation. The heatmap also providesattention clues for face tracking. The face tracking task also saves us theface detection procedure for each frame, which also significantly boost thereal-time capability for video-based tasks. We experimentally validate ATPN onfour benchmark datasets, WFLW, 300VW, WIDER Face and 300W-LP. The experimentalresults demonstrate that it achieves better performance with much lessparameters and lower computational complexity compared to other light models.

An Efficient Multitask Neural Network for Face Alignment, Head Pose Estimation and Face Tracking