3D-Guided Frontal Face Generation for Pose-Invariant Recognition

Although deep learning techniques have achieved extraordinary accuracy in recognizing human faces, the pose variances of images captured real-world scenarios still hinder reliable model appliance. To mitigate this gap, we propose to recognize faces via generation frontal face with a 3D -Guided Deep P ose- I nvariant Face Recognition M odel (3D-PIM) consisted simulator and refiner module. The employs Morphable Model (3D MM) fit shape appearance features recover primary less training data. further enhances image realism on both global facial structure local details adversarial training, while keeping discriminative identity information consistent original images. An Adaptive Weighting (AW) metric is then adopted leverage complimentary from recovered profile obtain credible similarity scores for recognition. Extended experiments verify superiority proposed “recognition generation” framework over state-of-the-art.