TR2018-116

Verification of Very Low-Resolution Faces Using An Identity-Preserving Deep Face Super-resolution Network


    •  Ataer-Cansizoglu, E., Jones, M.J., Zhang, Z., Sullivan, A., "Verification of Very Low-Resolution Faces Using An Identity-Preserving Deep Face Super-resolution Network", arXiv, August 2018.
      BibTeX arXiv
      • @article{Ataer-Cansizoglu2018aug,
      • author = {Ataer-Cansizoglu, Esra and Jones, Michael J. and Zhang, Ziming and Sullivan, Alan},
      • title = {Verification of Very Low-Resolution Faces Using An Identity-Preserving Deep Face Super-resolution Network},
      • journal = {arXiv},
      • year = 2018,
      • month = aug,
      • url = {https://arxiv.org/abs/1903.10974}
      • }
  • MERL Contacts:
  • Research Areas:

    Artificial Intelligence, Computer Vision, Machine Learning

Face super-resolution methods usually aim at producing visually appealing results rather than preserving distinctive features for further face identification. In this work, we propose a deep learning method for face verification on very low-resolution face images that involves identity-preserving face super-resolution with an extreme upscaling factor of 8. Our framework includes a super-resolution network and a feature extraction network. We train a VGG-based deep face recognition network [1] to be used as feature extractor. Our super-resolution network is trained to minimize the feature distance between the high resolution ground truth image and the super-resolved image, where features are extracted using our pre-trained feature extraction network. We carry out experiments on FRGC, Multi-PIE, LFW-a, and MegaFace datasets to evaluate our method in controlled and uncontrolled settings. The results show that the presented method outperforms conventional superresolution methods in low-resolution face verification.