检索结果-内蒙古大学图书馆

QuantFace: Towards Lightweight Face Recognition by Synthetic Data Low-bit Quantization

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Boutros, Fadi Damer, Naser Kuijper, Arjan Fraunhofer Institute for Computer Graphics Research IGD Darmstadt Germany Department of Computer Science TU Darmstadt Darmstadt Germany

Deep learning-based face recognition models follow the common trend in deep neural networks by utilizing full-precision floating-point networks with high computational costs. Deploying such networks in use-cases constrained by computational requirements is often infeasible due to the large memory required by the full-precision model. Previous compact face recognition approaches proposed to design special compact architectures and train them from scratch using real training data, which may not be available in a real-world scenario due to privacy concerns. We present in this work the QuantFace solution based on low-bit precision format model quantization. QuantFace reduces the required computational cost of the existing face recognition models without the need for designing a particular architecture or accessing real training data. QuantFace introduces privacy-friendly synthetic face data to the quantization process to mitigate potential privacy concerns and issues related to the accessibility to real training data. Through extensive evaluation experiments on seven benchmarks and four network architectures, we demonstrate that QuantFace can successfully reduce the model size up to 5x while maintaining, to a large degree, the verification performance of the full-precision model without accessing real training datasets. All training codes are publicly available 1 © 2022, CC BY-NC-ND.

关键词： Face recognition

Towards Explaining Demographic Bias through the Eyes of Face Recognition Models

学校读者我要写书评

暂无评论

Towards Explaining Demographic Bias through the Eyes of Face...

IEEE International Joint Conference on Biometrics (IJCB)

作者： Biying Fu Naser Damer Fraunhofer Institute for Computer Graphics Research IGD Darmstadt Germany Department of Computer Science TU Darmstadt Darmstadt Germany

ISBN: (纸本)9781665463959

Biases inherent in both data and algorithms make the fairness of widespread machine learning (ML)-based decision-making systems less than optimal. To improve the trustfulness of such ML decision systems, it is crucial to be aware of the inherent biases in these solutions and to make them more transparent to the public and developers. In this work, we aim at providing a set of explainability tool that analyse the difference in the face recognition models' behaviors when processing different demographic groups. We do that by leveraging higher-order statistical information based on activation maps to build explainability tools that link the FR models' behavior differences to certain facial regions. The experimental results on two datasets and two face recognition models pointed out certain areas of the face where the FR models react differently for certain demographic groups compared to reference groups. The outcome of these analyses interestingly aligns well with the results of studies that analyzed the anthropometric differences and the human judgment differences on the faces of different demographic groups. This is thus the first study that specifically tries to explain the biased behavior of FR models on different demographic groups and link it directly to the spatial facial features. The code is publicly available here 1 1 https://***/fbiying87/***.

关键词： Analytical models Machine learning algorithms Codes Face recognition Decision making Machine learning Behavioral sciences

Unsupervised Face Morphing Attack Detection via Self-paced Anomaly Detection

学校读者我要写书评

暂无评论

Unsupervised Face Morphing Attack Detection via Self-paced A...

IEEE International Joint Conference on Biometrics (IJCB)

作者： Meiling Fang Fadi Boutros Naser Damer Fraunhofer Institute for Computer Graphics Research IGD Darmstadt Germany Department of Computer Science TU Darmstadt Darmstadt Germany

ISBN: (纸本)9781665463959

The supervised-learning-based morphing attack detection (MAD) solutions achieve outstanding success in dealing with attacks from known morphing techniques and known data sources. However, given variations in the morphing attacks, the performance of supervised MAD solutions drops significantly due to the insufficient diversity and quantity of the existing MAD datasets. To address this concern, we propose a completely unsupervised MAD solution via self-paced anomaly detection (SPL-MAD) by leveraging the existing large-scale face recognition (FR) datasets and the unsupervised nature of convolutional autoencoders. Using general FR datasets that might contain unintentionally and unlabeled manipulated samples to train an autoencoder can lead to a diverse reconstruction behavior of attack and bona fide samples. We analyze this behavior empirically to provide a solid theoretical ground for designing our unsupervised MAD solution. This also results in proposing to integrate our adapted modified self-paced learning paradigm to enhance the reconstruction error separability between the bona fide and attack samples in a completely unsupervised manner. Our experimental results on a diverse set of MAD evaluation datasets show that the proposed unsupervised SPL-MAD solution outperforms the overall performance of a wide range of supervised MAD solutions and provides higher generalizability on unknown attacks. Training codes and pre-trained models are publicly released 1 1 https://***/meilfang/SPL-MAD.

关键词： Training Convolutional codes Adaptation models Face recognition Soft sensors Training data Solids

Towards Explaining Demographic Bias through the Eyes of Face Recognition Models

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Fu, Biying Damer, Naser Fraunhofer Institute for Computer Graphics Research IGD Darmstadt Germany Department of Computer Science TU Darmstadt Darmstadt Germany

Biases inherent in both data and algorithms make the fairness of widespread machine learning (ML)-based decision-making systems less than optimal. To improve the trustfulness of such ML decision systems, it is crucial to be aware of the inherent biases in these solutions and to make them more transparent to the public and developers. In this work, we aim at providing a set of explainability tool that analyse the difference in the face recognition models’ behaviors when processing different demographic groups. We do that by leveraging higher-order statistical information based on activation maps to build explainability tools that link the FR models’ behavior differences to certain facial regions. The experimental results on two datasets and two face recognition models pointed out certain areas of the face where the FR models react differently for certain demographic groups compared to reference groups. The outcome of these analyses interestingly aligns well with the results of studies that analyzed the anthropometric differences and the human judgment differences on the faces of different demographic groups. This is thus the first study that specifically tries to explain the biased behavior of FR models on different demographic groups and link it directly to the spatial facial features. The code is publicly available here1 © 2022, CC BY.

关键词： Face recognition

Unsupervised Face Morphing Attack Detection via Self-paced Anomaly Detection

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Fang, Meiling Boutros, Fadi Damer, Naser Fraunhofer Institute for Computer Graphics Research IGD Darmstadt Germany Department of Computer Science TU Darmstadt Darmstadt Germany

关键词： Face recognition

Unsupervised Face Recognition using Unlabeled Synthetic Data

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Boutros, Fadi Klemt, Marcel Fang, Meiling Kuijper, Arjan Damer, Naser Fraunhofer Institute for Computer Graphics Research IGD Darmstadt Germany Department of Computer Science TU Darmstadt Darmstadt Germany

Over the past years, the main research innovations in face recognition focused on training deep neural networks on large-scale identity-labeled datasets using variations of multi-class classification losses. However, many of these datasets are retreated by their creators due to increased privacy and ethical concerns. Very recently, privacy-friendly synthetic data has been proposed as an alternative to privacy-sensitive authentic data to comply with privacy regulations and to ensure the continuity of face recognition research. In this paper, we propose an unsupervised face recognition model based on unlabeled synthetic data (USynthFace). Our proposed USynthFace learns to maximize the similarity between two augmented images of the same synthetic instance. We enable this by a large set of geometric and color transformations in addition to GAN-based augmentation that contributes to the USynthFace model training. We also conduct numerous empirical studies on different components of our USynthFace. With the proposed set of augmentation operations, we proved the effectiveness of our USynthFace in achieving relatively high recognition accuracies using unlabeled synthetic data. The training code and pretrained model are publicly available under https://***/fdbtrs/Unsupervised-Face-Recognition-using-Unlabeled-Synthetic-Data. © 2022, CC BY-NC-SA.

关键词： Face recognition

Detecting Counterfeit Indian Currency A Comparative Analysis of Machine Learning Algorithms and Image Processing

学校读者我要写书评

暂无评论

Detecting Counterfeit Indian Currency A Comparative Analysis...

International Conference on Computing and Networking Technology (ICCNT)

作者： M R V Vyshnavi N.L. Varshitha Konda V. S. Harshith Kumar Tripty Singh Mansi Sharma Suman Chatterjee Department of Computer Science Engineering Amrita School of Computing Bengaluru India Graphics Era (Deemed to be university) Dehradun India Department of Mechanical Engineering Kongju National University Cheonan India

ISBN: (数字)9798350370249

ISBN: (纸本)9798350370270

The degree of secured currency is one of the key areas that makes its contribution in solving the problem of economic insecurity, highlights the most important. Even as it can be quite strained to actually identify which precise time frame this fake process was prepared in, complex fakes render the use of validation techniques quite challenging. This paper has aimed at displaying and evaluating the overall outcome of the different strategies of image, and the acquired classifiers in the identification of the fake FI currency. The following are the subsequent of the other operation that was applied in order to achieve the objectives of the intended work; Image capture This is whereby the images to be used are taken by the appropriate camera or scanner. Others used include the Naïve Bayes, Support Vector Classifier and Gradient Descent and Artificial Neural Network in this paper. As in the conclusion, with the help of the mathematical formulation, it can be confirmed that in fact, within an experiment, Neural Networks do have a higher level of optimization, much more than an average image processing if it is provoked to identify fake notes. In other words if used and implemented with the best of the machine learning approach and best of the image processing method then solitaire for detection of the counterfeit currency can be made more dependable than all the combined methods stated above. The objective of this study is to build and compare the performance of different machine learning algorithms and image processing methodologies for the identification of fake Indian currency notes.

关键词： Economics Training Machine learning algorithms Machine learning Artificial neural networks Vectors Real-time systems Bayes methods Reliability Currencies

Minimum Partition into Plane Subgraphs: The CG:SHOP Challenge 2022

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Fekete, Sándor P. Keldenich, Phillip Krupke, Dominik Schirra, Stefan Department of Computer Science TU Braunschweig Germany Department for Simulation and Graphics OvGU Magdeburg Germany

We give an overview of the 2022 Computational Geometry Challenge targeting the problem Minimum Partition into Plane Subsets, which consists of partitioning a given set of line segments into a minimum number of non-cro... 详细信息

关键词： Computational geometry

Evaluation of importance for condition attributes based on quality of decision reducts

学校读者我要写书评

暂无评论

Procedia computer science 2022年 207卷 2144-2153页

作者： Urszula Stańczyk Department of Graphics Computer Vision and Digital Systems Faculty of Automatic Control Electronics and Computer Science Silesian University of Technology Akademicka 2A 44-100 Gliwice Poland

Relative or decision reducts belong with mechanisms dedicated to feature selection, and they are embedded in rough set approach to data processing. Algorithms for reduct construction typically aim at dimensionality reduction aspect, searching for smallest reducts, which are considered as the most advantageous from the point of view of knowledge representation. However, classifiers build on reduced data models, based on reducts, can significantly vary in performance. Therefore, to ensure quality of predictions, other characteristics of reducts, apart from their cardinalities, need to be taken into account. The paper presents research in which estimation of reduct quality through their characteristics was reflected in calculation of the proposed weighting factors leading to attribute rankings. These rankings were next employed in the process of filtering decision rules, inferred by classic rough set approach. Constructed rule-based classifiers were applied in the stylometric domain to solve a task of authorship attribution.

关键词： Decision reduct Rough set theory Ranking Weighting Feature selection Rule filtering