Deep face recognition: A survey
TL;DR: A comprehensive review of the recent developments on deep face recognition can be found in this paper, covering broad topics on algorithm designs, databases, protocols, and application scenes, as well as the technical challenges and several promising directions.
About: This article is published in Neurocomputing.The article was published on 2021-03-14 and is currently open access. It has received 353 citations till now. The article focuses on the topics: Deep learning & Feature extraction.
Citations
More filters
TL;DR: Current state-of-the-art privacy mechanisms provide good protection in principle but there is no compelling one-size-fits-all privacy approach, which leads to further questions regarding the practicality of these mechanisms, which are presented in the form of seven thought-provoking propositions.
Abstract: The present information age is characterized by an ever-increasing digitalization. Smart devices quantify our entire lives. These collected data provide the foundation for data-driven services called smart services. They are able to adapt to a given context and thus tailor their functionalities to the user’s needs. It is therefore not surprising that their main resource, namely data, is nowadays a valuable commodity that can also be traded. However, this trend does not only have positive sides, as the gathered data reveal a lot of information about various data subjects. To prevent uncontrolled insights into private or confidential matters, data protection laws restrict the processing of sensitive data. One key factor in this regard is user-friendly privacy mechanisms. In this paper, we therefore assess current state-of-the-art privacy mechanisms. To this end, we initially identify forms of data processing applied by smart services. We then discuss privacy mechanisms suited for these use cases. Our findings reveal that current state-of-the-art privacy mechanisms provide good protection in principle, but there is no compelling one-size-fits-all privacy approach. This leads to further questions regarding the practicality of these mechanisms, which we present in the form of seven thought-provoking propositions.
6 citations
TL;DR: The authors proposed to automatically code campaign advertisement videos by applying state-of-the-art machine learning methods, extracting various audio and image features from each video file and show that their machine coding is comparable to human coding for many variables of the WMP datasets.
Abstract:
Video advertisements, either through television or the Internet, play an essential role in modern political campaigns. For over two decades, researchers have studied television video ads by analyzing the hand-coded data from the Wisconsin Advertising Project and its successor, the Wesleyan Media Project (WMP). Unfortunately, manually coding more than a hundred of variables, such as issue mentions, opponent appearance, and negativity, for many videos is a laborious and expensive process. We propose to automatically code campaign advertisement videos. Applying state-of-the-art machine learning methods, we extract various audio and image features from each video file. We show that our machine coding is comparable to human coding for many variables of the WMP datasets. Since many candidates make their advertisement videos available on the Internet, automated coding can dramatically improve the efficiency and scope of campaign advertisement research. Open-source software package is available for implementing the proposed methodology.
6 citations
TL;DR: A new label‐only membership inference attack scheme targeted at machine unlearning to eliminate the dependence on posteriors is proposed and achieves high inference accuracy (measured by AUC) in label‐ only settings.
Abstract: Machine unlearning is the process through which a deployed machine learning model is enforced to forget about some of its training data items. It normally generates two machine learning models, the original model and the unlearned model, indicating training results before and after data items are deleted. However, recent studies find that machine unlearning is vulnerable to membership inference attacks—as the directivity of training and nontraining data (i.e., data items in the training set have high posterior probabilities), the attackers can utilize this property to infer whether an item has been used for original model training. Nevertheless, such attacks are incapable in label‐only settings, in which the attackers are infeasible to get the posteriors. In this paper, we propose a new label‐only membership inference attack scheme targeted at machine unlearning to eliminate the dependence on posteriors. Our heuristic is that injected turbulence on candidate samples will present different behaviors for training and nontraining data. Thus, in our scheme, the attacker iteratively query on the original/unlearned models and inject turbulence to change their predicting labels; it determines whether an item is having‐been‐delated by observing the disturbance amplitude. Extensive experiments (i.e., on MNIST, CIFAR10, CIFAR100, and STL10 data sets) show that our method achieves high inference accuracy (measured by AUC) in label‐only settings, for example, AUC = 0.96 for MNIST data set. Besides, we analyze the existing countermeasures in mitigating inference attacks and find that our scheme can bypass most of them.
6 citations
TL;DR: A Light and Fast Face Detector with an Ommateum Structure (OS-LFFD) is proposed, which reduces the number of model parameters (8 M), which makes it much smaller than most face detectors, and can considerably balance the accuracy and running speed.
Abstract: Face detection has been deployed on edge devices as the basis for face applications, but the devices cannot store large-scale models and have low computing power. The existing anchor-based face detection schemes cannot cover face images over a continuous size range, and their performance is not satisfactory. Obviously, good performances are accompanied by increased storage and lower speed. We find that the feature points in different layers correspond to a specific size range of RFs (receptive fields). According to the survey, the predictable range of RFs with the same size is the face on a continuous scale. Therefore, we argue that RFs are inherent anchors. A Light and Fast Face Detector with an Ommateum Structure (OS-LFFD) is proposed in this paper. By analyzing the correlation between the effective receptive field (ERF) and face sizes, a 4-branch network is designed to cover the objective range of face sizes. Each branch involves an ommateum block with a similar structure and shared parameters. It reduces the number of model parameters (8 M), which makes it much smaller than most face detectors. Experiments on the popular benchmarks WIDER FACE and FDDB using multiple hardware platforms demonstrate that the proposed scheme can considerably balance the accuracy and running speed.
5 citations
DOI•
24 Nov 2021TL;DR: In this paper, the authors presented the limitations of deep learning based methods in face recognition tasks, and proposed two methods to improve the performance of face authentication system by using Open-Set concepts.
Abstract: In everyday life, authentication is an indispensable process of human activities. Bio-metric authentication system is one of the effective solutions, because it uses human-based features, instead of other traditional features, such as pin, password, etc. However, to apply a face authentication system in practical applications, we need to ensure that the system must not try to recognize the face of an unknown person into known categories, meaning we need to reject faces of unknown people in our application. In this paper, we present the limitations of recent Deep Learning based methods in Face Recognition tasks. We then propose two methods helping Face Recognition system have the ability to reject faces from unknown people by using Open-Set concepts. We conduct the experiments on a subset of CASIA-WebFace dataset, with a train set that includes 7000 images of 100 known people and a test set that includes both known and unknown people. Without rejecting unknown faces, the regular face recognition, i.e. the baseline method, yields the accuracy of only 45.9%, as the method tries to classify all face photos into known classes. Our proposed methods, which are combined deep network of Facenet system with recent Open Set methods, are called Learning Placeholder on Facenet (P-Facenet) and Facenet with OpenMax (O-Facenet). They achieve the accuracy of 83.6% and 88.5% respectively. This is a potential approach for authentication with face recognition to decrease the error rate of the model when recognizing faces of unknown people in the wild.
5 citations
References
More filters
27 Jun 2016
TL;DR: In this article, the authors proposed a residual learning framework to ease the training of networks that are substantially deeper than those used previously, which won the 1st place on the ILSVRC 2015 classification task.
Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.
123,388 citations
Proceedings Article•
03 Dec 2012TL;DR: The state-of-the-art performance of CNNs was achieved by Deep Convolutional Neural Networks (DCNNs) as discussed by the authors, which consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax.
Abstract: We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.
73,978 citations
Proceedings Article•
04 Sep 2014TL;DR: This work investigates the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting using an architecture with very small convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers.
Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.
55,235 citations
07 Jun 2015
TL;DR: Inception as mentioned in this paper is a deep convolutional neural network architecture that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14).
Abstract: We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.
40,257 citations
08 Dec 2014
TL;DR: A new framework for estimating generative models via an adversarial process, in which two models are simultaneously train: a generative model G that captures the data distribution and a discriminative model D that estimates the probability that a sample came from the training data rather than G.
Abstract: We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to ½ everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.
38,211 citations