Abstract
We propose an ID Preserving Face Super-Resolution Generative Adversarial Networks (IP-FSRGAN) to reconstruct realistic super-resolution face images from low-resolution ones. Inspired by the success of generative adversarial networks (GAN), we introduce a novel ID preserving module to help the generator learn to infer the facial details and synthesize more realistic super-resolution faces. Our method produces satisfactory visual results and also quantitatively outperforms state-of-the-art super-resolution methods on the face datasets including CASIA-Webface, CelebA, and LFW datasets under the metrics of PSNR, SSIM, and cosine similarity. In addition, we propose a framework to apply IP-FSRGAN model to address the face verification task on low-resolution face images. The synthesized $4\times $ super-resolution faces achieve a verification accuracy of 97.6%, improved from 92.8% of low resolution faces. We also prove by experiments that the proposed IP-FSRGAN model demonstrates excellent robustness under different downsample scaling factors and extensibility to various face verification models.
Highlights
Recent advancements in deep neural networks and generative adversarial networks (GANs) have brought tremendous successes in the super-resolution task, which reconstructs high-resolution images from low-resolution images [1]
We propose an ID Preserving Face Super-Resolution Generative Adversarial Networks (IP-FSRGAN) to address the distortion problem and reconstruct realistic high-resolution faces
We prove by experiments that the proposed IP-FSRGAN could improve face verification on low-resolution face image
Summary
Recent advancements in deep neural networks and generative adversarial networks (GANs) have brought tremendous successes in the super-resolution task, which reconstructs high-resolution images from low-resolution images [1]. We propose an ID Preserving Face Super-Resolution Generative Adversarial Networks (IP-FSRGAN) to address the distortion problem and reconstruct realistic high-resolution faces. We prove by experiments that the proposed IP-FSRGAN could improve face verification on low-resolution face image. The main contributions of this paper are three folds: Firstly, we propose IP-FSRGAN, the first to integrate id preserving loss to the setting of GANs, for face super-resolution and design the corresponding training strategy. We prove by experiments that the proposed IP-FSRGAN could generate realistic HR faces and improve the performance of face verification and recognition. The GANs based models usually lead to blurring and distortion due to information missing of the low-resolution faces This limits its application in some advanced facial tasks such as face recognition. We empirically determine the optimal λ, γ , η and ξ
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
