Joint face detection and Facial Landmark Localization using graph match and pseudo label

Highlights：

• A real-time framework for joint face detection and facial landmark localization is proposed.

• The synergy between the face detection and facial landmark localization is exploited. A fully convolutional network to predict the location of facial landmarks and face regions is designed.

• The cluster assumption in facial landmark localization is observed to guide a progressively pseudo labeling training method. The method not only eliminates the harmful effect caused by inaccurate/noisy annotations, but also unifies the exact, inexact and coarse-grained labels.

• Two graph matching algorithms without learnable parameters are proposed to complete the bottom-up face assembly after model inference.

• Extensive experiments show that our approaches achieve state-of-the-art level on face detection and facial landmark localization in the different datasets, both in accuracy and runtime.

摘要

•A real-time framework for joint face detection and facial landmark localization is proposed.•The synergy between the face detection and facial landmark localization is exploited. A fully convolutional network to predict the location of facial landmarks and face regions is designed.•The cluster assumption in facial landmark localization is observed to guide a progressively pseudo labeling training method. The method not only eliminates the harmful effect caused by inaccurate/noisy annotations, but also unifies the exact, inexact and coarse-grained labels.•Two graph matching algorithms without learnable parameters are proposed to complete the bottom-up face assembly after model inference.•Extensive experiments show that our approaches achieve state-of-the-art level on face detection and facial landmark localization in the different datasets, both in accuracy and runtime.