Deep autoencoders for attribute preserving face de-identification

The mass availability of mobile devices equipped with cameras has lead to increased public privacy concerns in recent years. Face de-identification is a necessary first step towards anonymity preservation, and can be trivially solved by blurring or concealing detected faces. However, such naive privacy protection methods are both ineffective and unsatisfying, producing a visually unpleasant result. In this paper, we tackle face de-identification using Deep Autoencoders, by finetuning the encoder to perform face de-identification. We present various methods to finetune the encoder in both a supervised and unsupervised fashion to preserve facial attributes, while generating new faces which are both visually and quantitatively different from the original ones. Furthermore, we quantify the realism and naturalness of the resulting faces by introducing a diversity metric to measure the distinctiveness of the new faces. Experimental results show that the proposed methods can generate new faces with different person identity labels, while maintaining the facelike nature and diversity of the input face images.     

应用于人脸识别的监督局部邻域保持嵌入算法

提出了一种应用于人脸识别的监督线性维数约简 算法。首先引入图像距离度量方法以确定人脸数据 之间的相似程度,之后将训练样本的类标先验信息融入到邻域保持嵌入(NPE,neighborhood preserving embedding)算法的目标函 数中,使得降维后的嵌入空间的投影数据呈多流形分布,不仅最优保持了样本空间的局部几 何结构,同时各类样本 投影的类内散度最小化,类间散度最大化,增大了各类数据分布之间的间隔,提高了嵌入空 间的辨别能力。在Extended Yale B和CMU PIE两个开放人脸数据库上进行了识别实验,结果表明,本文算法取得了很好 的识别效果。     

Face model fitting on video sequences based on incremental virtual active appearance model

It has been proved that the performance of a person-specific active appearance model (AAM) built to model the appearance variation of a single person across pose, illumination, and expression is substantially better than the performance of a generic AAM built to model the appearance variation of many faces. However, it is not practical to build a personal AAM before tracking an unseen subject. A virtual person-specific AAM is proposed to tackle the problem. The AAM is constructed from a set of virtual personal images with different poses and expressions which are synthesized from the annotated first frame via regressions. To preserve personal facial details on the virtual images, a poison fusion strategy is designed and applied to the virtual facial images generated via bilinear kernel ridge regression. Furthermore, the AAM subspace is sequentially updated during tracking based on sequential Karhunen–Loeve algorithm, which helps the AAM adaptive to the facial context variation. Experiments show the proposed virtual personal AAM is robust to facial context changes during tracking, and outperforms other state-of-the-art AAM on facial feature tracking accuracy and computation cost.     

Real-time Stereo Face Recognition by Fusing Appearance and Depth Fisherfaces

Reported 3D face recognition techniques assume the use of active 3D measurement for 3D facial capture. However, active method employ structured illumination (structure projection, phase shift, gray-code demodulation, etc) or laser scanning, which is not desirable in many applications. A major problem of using passive stereo is its lower 3D face resolution and thus no passive method for 3D face recognition has been reported. In this paper, a real-time passive stereo face recognition system is presented. Entire face detection, tracking, pose estimation and face recognition are investigated. We used SRI Stereo engine that outputs sub-pixel disparity automatically. An investigation is carried out in combining 3D and 2D information for face recognition. The straightforward two-stage principal component analysis plus linear discriminant analysis is carried out in appearance and depth face images respectively. A probe face is identified using sum of the weighted appearance and depth linear discriminant distances. We investigate the complete range of linear combinations to reveal the interplay between these two paradigms. The improvement of the face recognition rate using this combination is verified. The recognition rate by the combination is higher than that of either appearance alone or depth alone. We then discuss the implementation of the algorithm on a stereo vision system. A hybrid face and facial features detection/tracking approach is proposed which collects near-frontal views for face recognition. Our face detection/tracking approach automatically initializes without user intervention and can be re-initialized automatically if the tracking of the 3D face pose is lost. The experiments include two parts. Firstly, the performance of the proposed algorithm is verified on XM2VTS database; Secondly, the algorithm is demonstrated on a real-time stereo vision system. It is able to detect, track and recognize a person while walking toward a stereo camera.

基于两空间核鉴别分析的人脸识别

指出了几种线性鉴别分析方法在处理小样本人脸识别问题时存在的不足,结合核方法的思想,提出了一种基于两空间核鉴别分析的人脸识别方法.首先使用KPCA方法在核变换后的特征空间中对样本进行处理;进面将变换后的类内散布矩阵分成非零空间和零空间进行鉴别向量确定和鉴别特征提取,最后将得到的两种鉴别特征融合,从而使用最近邻法进行分类....     

3DFP-FCGAN: Face completion generative adversarial network with 3D facial prior

Face completion is a domain-specific image inpainting problem. Most existing face completion methods fail to synthesize fine-grained facial structures due to the undifferentiated treatment of face images and other scene images. To handle this problem, we propose an end-to-end deep generative model based approach which makes full use of the facial prior knowledge, including 2D facial geometry priors from facial parsing maps and landmarks, as well as the 3D depth prior. We adopt a coarse-to-fine inpainting framework where the 2D facial geometry priors based on coarse faces are extracted to guide the refinement network for better planar facial textures and structures. Moreover, a novel 3D regularized reconstruction loss is proposed for the enhancement of the stereo perception of generated faces. Experimental results on two large-scale benchmarks CelebA and CelebA-HQ show that our method significantly outperforms the state-of-the-arts in generating more visually realistic and pleasing faces. Code is available at .     

耳朵信息对侧脸检测的影响研究

如何准确地检测和定位图像中的人脸是人脸检测领域的关键问题.为了进一步提高人脸检测器的性能,常见的方法是增加训练数据集或采用更鲁棒的人脸特征表示,而训练人脸检测器的一个基础工作是：为训练图像中的人脸标注边界框.但标注的人脸边界框是否应该包含耳朵信息,以及对训练出的侧脸检测器性能的影响尚未被研究.本文的实验结果表明：在侧脸数据集上训练人脸检测器时,如果人脸边界框包含耳朵信息,基于DPM（Deformable Parts Model）方法训练得到的侧脸检测器使侧脸检测的准确率降低1.9%,召回率提高6.3%.而基于Viola&Jones和Fast R-CNN方法训练得到的侧脸检测器使准确率分别提高6.8%和4.4%,召回率分别提高14.9%和12.9%.这说明包含耳朵信息训练出的侧脸检测模型,有助于提高侧脸检测率.     

一种快速彩色图像中复杂背景下人脸检测方法

本文提出了一种快速的彩色图像中复杂背景下人脸检测方法.该方法首先利用进化Agent计算对彩色图像中与人的肤色相似的像素进行聚类和区域分割.然后利用小波分解对每一个侯选区域进行人脸特征分析,如所检测到的区域特征分布相似于某一预先定义的人脸模型,则确认该区域代表人脸.实验结果表明,该方法不仅速度快、效率高,而且正确检测率高.     

Generalized face super-resolution.

Existing learning-based face super-resolution (hallucination) techniques generate high-resolution images of a single facial modality (i.e., at a fixed expression, pose and illumination) given one or set of low-resolution face images as probe. Here, we present a generalized approach based on a hierarchical tensor (multilinear) space representation for hallucinating high-resolution face images across multiple modalities, achieving generalization to variations in expression and pose. In particular, we formulate a unified tensor which can be reduced to two parts: a global image-based tensor for modeling the mappings among different facial modalities, and a local patch-based multiresolution tensor for incorporating high-resolution image details. For realistic hallucination of unregistered low-resolution faces contained in raw images, we develop an automatic face alignment algorithm capable of pixel-wise alignment by iteratively warping the probing face to its projection in the space of training face images. Our experiments show not only performance superiority over existing benchmark face super-resolution techniques on single modal face hallucination, but also novelty of our approach in coping with multimodal hallucination and its robustness in automatic alignment under practical imaging conditions.     

基于小波分解和支持向量机的准正面人脸识别方法

基于小波分解提取人脸特征技术和多分类支持向量机模型，提出了一种新的准正面人脸识别算法。小波分解提取人脸特征具有对表情变化不敏感的特点；支持向量机作为分类器被认为具有很高的推广(generalization)性能，无需先验知识。在所提出的算法中，首先对训练图像进行预处理，然后使用小波分解方法对人脸图像进行特征提取，用所提取的人脸特征向量训练多分类支持向量机模型，最后用训练好的支持向量机进行人脸识别。利用ORL人脸图像库对该算法的实验测试结果，以及与其它人脸识别方法的比较结果表明了该算法在识别性能方面的优越性。     

Mutual information maximizing GAN inversion for real face with identity preservation

Recent generative adversarial networks (GANs) have yielded remarkable performance in face image synthesis. GAN inversion embeds an image into the latent space of a pretrained generator, enabling it to be used for real face manipulation. However, current inversion approaches for real faces suffer the dilemma of initialization collapse and identity loss. In this paper, we propose a hierarchical GAN inversion for real faces with identity preservation based on mutual information maximization. We first use a facial domain guaranteed initialization to avoid the initialization collapse. Furthermore, we prove that maximizing the mutual information between inverted faces and their identities is equivalent to minimizing the distance between identity features from inverted and original faces. Optimization for real face inversion with identity preservation is implemented on this mutual information-maximizing constraint. Extensive experimental results show that our approach outperforms state-of-the-art solutions for inverting and editing real faces, particularly in terms of face identity preservation.     

Bilinear factorisation for facial expression analysis and synthesis

The paper addresses the issue of face representations for facial expression recognition and synthesis. In this context, a global active appearance model is used in conjunction with two bilinear factorisation models to separate expression and identity factors from the global appearance parameters. Although active appearance models and bilinear modelling are not new concepts, the paper's main contribution consists in combining both techniques to improve facial expression recognition and synthesis (control). Indeed, facial expression recognition is performed through linear discriminant analysis of the global appearance parameters extracted by active appearance model search. Results are compared to ones obtained for the same training and test images using classification of the expression factors extracted by bilinear factorisation. This experiment highlights the advantages of bilinear factorisation. Finally, it is proposed to exploit bilinear factorisation to synthesise facial expressions through replacement of the extracted expression factors. This yields very interesting synthesis performances in terms of visual quality of the synthetic faces. Indeed, synthetic open mouth reconstruction, either with or without teeth appearing, is of better quality than with classical linear-regression-based synthesis.     

基于图像隶属度的主分量分析人脸识别算法

提出了基于图像隶属度的主分量人脸识别算法.该算法首先用小波变换对人脸图像进行小波分解,形成低频小波子图,然后用主分量分析法构造特征脸子空间.计算训练样本和待测样本在人脸特征空间中的投影向量间的距离.引入图像隶属函数,作为识别分类器进行人脸识别.针对ORL人脸库的实验结果表明该方法具有良好的识别分离能力.     

平面旋转人脸检测与特征定位方法研究

提出了一种基于角点检测、AdaBoost算法和C-V方法的平面旋转人脸检测及特征定位方法.方法首先根据AdaBoost算法训练样本得到脸、眼、鼻、嘴4个检测器;然后以角点作为眼睛的候选点,枚举任意两个角点构造可能的人脸区域,并在区域内运用人脸检测器进行检测;接着利用眼、鼻、嘴检测器检测出人脸特征所在的矩形区域;最后利用C-V方法从各个特征区域中分割出人脸特征的轮廓,进而得到人脸关键特征点的位置.在CMU平面旋转测试集上的检测率为94.6%,误报24个,提取出的特征点位置准确.实验结果表明方法是有效的.     

基于二维Fisher线性判别的人脸耳组合识别

针对人脸易受到年龄、表情等影响,提出了脸和耳相结合的组合识别方法。利用二维Fisher线性判别(2DFLD)方法分别进行了脸、耳图像层和特征层的组合识别。在北京科技大学人耳库和ORL人脸库上进行实验,结果表明,图像层组合和特征层组合的识别率分别为97.5%、95.0%,分别比人脸识别提高了12.5%和10.0%,比人耳识别提高了5.0%和2.5%;与同样应用于组合识别的主成分分析(PCA)、二维PCA(2DPCA)比较,也取得了较好识别效果。这说明,多生物特征组合识别是一种有效的识别方法。     

具有学习功能的自动人脸识别

人脸识别是模式识别领域中一个相当困难而又有理论意义和实际价值的研究课题。传统的基于K－L变换的自动人脸识别方法,不用过多地考虑人脸的局部特征,利用特征脸方法进行识别,取得了一定的。但是,人脸作为一个特殊的场景,脸像会受年龄、心情、拍摄角度、光照条件、发饰等因素影响,所成图像存在差异。传统的基于K－L变换的自动人脸识别方法不能很好地克服这些畸变的影响。文中就主成分分析方法引入人脸识别,模拟人脸脸像的各种变化,事先对脸像做相应的变化,产生一系列变形脸。然后对变形脸进行主成分分析,提取它们的主成分。最后应用遗传算法选择最优特征向量构造子空间,提出一种能抗御一定脸像变化的人脸识别方法,并运用该方法进行了实验。实验结果证明了该方法的可行性和良好的抗畸变能力。     

多姿态、光照不均人脸图像的子空间人脸识别方法研究

针对人脸识别中存在的光照不均匀问题,提出了一种预处理链技术,能达到很好的光照补偿效果。为了提高多姿态、多表情、多细节人脸图像的人脸识别率,设计了一种将最近邻分类器与支持向量机相结合的分类算法（NN-SVM）,基于该分类算法提出了一种基于Gabor变换和NN-SVM的子空间人脸识别方法。在FERET和ORL两大人脸数据库中对所提方法进行性能评估,实验结果表明所提出方法能有效的解决人脸识别中光照不均匀问题,大大的提高人脸识别率,而且相比其他现存的人脸识别方法,所设计的方法具有更好、更稳定的识别效果。