Reconstruction-Based Disentanglement for Pose-Invariant Face Recognition

@article{Peng2017ReconstructionBasedDF,
  title={Reconstruction-Based Disentanglement for Pose-Invariant Face Recognition},
  author={Xi Peng and Xiang Yu and Kihyuk Sohn and Dimitris N. Metaxas and Manmohan Chandraker},
  journal={2017 IEEE International Conference on Computer Vision (ICCV)},
  year={2017},
  pages={1632-1641}
}
Deep neural networks (DNNs) trained on large-scale datasets have recently achieved impressive improvements in face recognition. [] Key Method Finally, we propose a new feature reconstruction metric learning to explicitly disentangle identity and pose, by demanding alignment between the feature reconstructions through various combinations of identity and pose features, which is obtained from two images of the same subject. Experiments on both controlled and in-the-wild face datasets, such as MultiPIE, 300WLP…
PoseFace: Pose-Invariant Features and Pose-Adaptive Loss for Face Recognition
TLDR
An efficient PoseFace framework is proposed which utilizes the facial landmarks to disentangle the poseinvariant features and exploits a pose-adaptive loss to handle the imbalance issue adaptively.
Attention-guided Progressive Mapping for Profile Face Recognition
TLDR
This paper proposes to learn the feature residual between the source pose and its nearby pose in a block-by-block fashion, and thus traversing to the feature space of a smaller pose by adding the learned residual.
Towards Large-Pose Face Frontalization in the Wild
TLDR
This work proposes a novel deep 3D Morphable Model (3DMM) conditioned Face Frontalization Generative Adversarial Network (GAN), termed as FF-GAN, to generate neutral head pose face images, which differs from both traditional GANs and 3DMM based modeling.
A 3D GAN for Improved Large-pose Facial Recognition
TLDR
A 3D morphable model is incorporated into the generator of a GAN in order to learn a nonlinear texture model from in-the-wild images, which allows generation of new, synthetic identities, and manipulation of pose, illumination and expression without compromising the identity.
Unsupervised Disentangling of Viewpoint and Residues Variations by Substituting Representations for Robust Face Recognition
TLDR
This paper proposes a novel disentangling framework through modeling three representations of identity, viewpoint, and residues of facial images which do not require supervision of the variations and achieves robust face recognition performance.
Adversarial Pose Regression Network for Pose-Invariant Face Recognitions
TLDR
An Adversarial Pose Regression Network (APRN) is proposed to extract poseinvariant identity representations by disentangling their pose variation in hidden feature maps and achieves comparable or even superior to the state-of-the-art on CFP, Multi-PIE, IJB-A and MegaFace datasets.
Deformable Face Net: Learning Pose Invariant Feature with Pose Aware Feature Alignment for Face Recognition
TLDR
Different from the existing methods, the DFN focuses on aligning features across different poses rather than frontalizing the input faces, and outperforms the state-of-the-art methods, especially on the datasets with large poses.
Pose-Guided Photorealistic Face Rotation
TLDR
This work focuses on flexible face rotation of arbitrary head poses, including extreme profile views, with a novel Couple-Agent Pose-Guided Generative Adversarial Network (CAPG-GAN) to generate both neutral and profile head pose face images.
...
...

References

SHOWING 1-10 OF 57 REFERENCES
Disentangled Representation Learning GAN for Pose-Invariant Face Recognition
TLDR
Quantitative and qualitative evaluation on both controlled and in-the-wild databases demonstrate the superiority of DR-GAN over the state of the art.
Towards Large-Pose Face Frontalization in the Wild
TLDR
This work proposes a novel deep 3D Morphable Model (3DMM) conditioned Face Frontalization Generative Adversarial Network (GAN), termed as FF-GAN, to generate neutral head pose face images, which differs from both traditional GANs and 3DMM based modeling.
Stacked Progressive Auto-Encoders (SPAE) for Face Recognition Across Poses
TLDR
The proposed method to learn pose-robust features by modeling the complex non-linear transform from the non-frontal face images to frontal ones through a deep network in a progressive way, termed as stacked progressive auto-encoders (SPAE).
Deep Learning Identity-Preserving Face Space
TLDR
This paper proposes a new learning based face representation: the face identity-preserving (FIP) features, a deep network that combines the feature extraction layers and the reconstruction layer that significantly outperforms the state-of-the-art face recognition methods.
Pose-Aware Face Recognition in the Wild
TLDR
A method to push the frontiers of unconstrained face recognition in the wild by using multiple pose specific models and rendered face images called Pose-Aware Models (PAMs), which achieve remarkably better performance than commercial products and surprisingly also outperform methods that are specifically fine-tuned on the target dataset.
DeepFace: Closing the Gap to Human-Level Performance in Face Verification
TLDR
This work revisits both the alignment step and the representation step by employing explicit 3D face modeling in order to apply a piecewise affine transformation, and derive a face representation from a nine-layer deep neural network.
Triplet probabilistic embedding for face verification and clustering
TLDR
This paper proposes an approach that couples a deep CNN-based approach with a low-dimensional discriminative embedding step, learned using triplet probability constraints to address the unconstrained face verification problem.
High-fidelity Pose and Expression Normalization for face recognition in the wild
TLDR
A High-fidelity Pose and Expression Normalization (HPEN) method with 3D Morphable Model (3DMM) which can automatically generate a natural face image in frontal pose and neutral expression and an inpainting method based on Possion Editing to fill the invisible region caused by self occlusion is proposed.
Face Alignment Across Large Poses: A 3D Solution
TLDR
3D Dense Face Alignment (3DDFA), in which a dense 3D face model is fitted to the image via convolutional neutral network (CNN), is proposed, and a method to synthesize large-scale training samples in profile views to solve the third problem of data labelling is proposed.
Multi-View Perceptron: a Deep Model for Learning Face Identity and View Representations
TLDR
This paper proposes a novel deep neural net, named multi-view perceptron (MVP), which can untangle the identity and view features, and in the meanwhile infer a full spectrum of multi- view images, given a single 2D face image.
...
...