12 sonuçlar
Arama Sonuçları
Listeleniyor 1 - 10 / 12
Yayın Unsupervised textile defect detection using convolutional neural networks(Elsevier Ltd, 2021-12) Koulali, Imane; Eskil, Mustafa TanerIn this study, we propose a novel motif-based approach for unsupervised textile anomaly detection that combines the benefits of traditional convolutional neural networks with those of an unsupervised learning paradigm. It consists of five main steps: preprocessing, automatic pattern period extraction, patch extraction, features selection and anomaly detection. This proposed approach uses a new dynamic and heuristic method for feature selection which avoids the drawbacks of initialization of the number of filters (neurons) and their weights, and those of the backpropagation mechanism such as the vanishing gradients, which are common practice in the state-of-the-art methods. The design and training of the network are performed in a dynamic and input domain-based manner and, thus, no ad-hoc configurations are required. Before building the model, only the number of layers and the stride are defined. We do not initialize the weights randomly nor do we define the filter size or number of filters as conventionally done in CNN-based approaches. This reduces effort and time spent on hyper-parameter initialization and fine-tuning. Only one defect-free sample is required for training and no further labeled data is needed. The trained network is then used to detect anomalies on defective fabric samples. We demonstrate the effectiveness of our approach on the Patterned Fabrics benchmark dataset. Our algorithm yields reliable and competitive results (on recall, precision, accuracy and f1-measure) compared to state-of-the-art unsupervised approaches, in less time, with efficient training in a single epoch and a lower computational cost.Yayın Integrating vendors into cooperative design practices(Taylor & Francis Ltd, 2009) Eskil, Mustafa Taner; Sticklen, JonThis paper describes a new approach to cooperative design using distributed, off-the-shelf design components. The ultimate goal is to enable assemblers to rapidly design their products and perform simulations using parts that are offered by a global network of suppliers. The obvious way to realise this goal would be to transfer desired component models to the client computer. However, in order to protect proprietary data, manufacturers are reluctant to share their design models without non-disclosure agreements, which can take in the order of months to put in place. Due to bandwidth limitations, it is also impractical to keep the models at the manufacturer site and do simulations by simple message passing. To deal with these impediments in e-commerce the modular distributed modelling (MDM) methodology is leveraged, which enables transfer of component models while hiding proprietary implementation details. MDM methodology with routine design (RD) methods are augmented to realise a platform (RD-MDM) that enables automatic selection of secured off-the-shelf design components over the Internet, integration of these components in an assembly, running simulations for design testing and publishing the approved product model as a secured MDM agent. This paper demonstrates the capabilities of the RD-MDM platform on a fuel cell-battery hybrid vehicle design example.Yayın Nearest neighbor weighted average customization for modeling faces(Springer, 2013-10) Abeysundera, Hasith Pasindu; Benli, Kristin Surpuhi; Eskil, Mustafa TanerIn this paper, we present an anatomically accurate generic wireframe face model and an efficient customization method for modeling human faces. We use a single 2D image for customization of the generic model. We employ perspective projection to estimate 3D coordinates of the 2D facial landmarks in the image. The non-landmark vertices of the 3D model are shifted using the translations of k nearest landmark vertices, inversely weighted by the square of their distances. We demonstrate on Photoface and Bosphorus 3D face data sets that the proposed method achieves substantially low relative error values with modest time complexity.Yayın Factored particle filtering with dependent and constrained partition dynamics for tracking deformable objects(Springer, 2014-10) Eskil, Mustafa TanerIn particle filtering, dimensionality of the state space can be reduced by tracking control (or feature) points as independent objects, which are traditionally named as partitions. Two critical decisions have to be made in implementation of reduced state-space dimensionality. First is how to construct a dynamic (transition) model for partitions that are inherently dependent. Second critical decision is how to filter partition states such that a viable and likely object state is achieved. In this study, we present a correlation-based transition model and a proposal function that incorporate partition dependency in particle filtering in a computationally tractable manner. We test our algorithm on challenging examples of occlusion, clutter and drastic changes in relative speeds of partitions. Our successful results with as low as 10 particles per partition indicate that the proposed algorithm is both robust and efficient.Yayın The routine design-modular distributed modeling platform for distributed routine design and simulation-based testing of distributed assemblies(Cambridge University Press, 2008-12-12) Eskil, Mustafa Taner; Sticklen, Jon; Radcliffe, ClarkIn this paper we describe a conceptual framework and implementation of a tool that supports task-directed, distributed routine design (RD) augmented with simulation-based design testing. In our research, we leverage the modular distributed modeling (MDM) methodology to simulate the interaction of design components in an assembly. The major improvement we have made in the RD methodology is to extend it with the capabilities of incorporating remotely represented off-the-shelf components in design and simulation-based testing of a distributed assembly. The deliverable of our research is the RD-MDM platform, which is capable of automatically selecting intellectually protected off the shelf design components over the Internet, integrating these components in an assembly, running simulations for design testing, and publishing the approved design without disclosing the proprietary information.Yayın Facial expression recognition based on anatomy(Academic Press Inc Elsevier Science, 2014-02) Eskil, Mustafa Taner; Benli, Kristin SurpuhiIn this study, we propose a novel approach to facial expression recognition that capitalizes on the anatomical structure of the human face. We model human face with a high-polygon wireframe model that embeds all major muscles. Influence regions of facial muscles are estimated through a semi-automatic customization process. These regions are projected to the image plane to determine feature points. Relative displacement of each feature point between two image frames is treated as an evidence of muscular activity. Feature point displacements are projected back to the 3D space to estimate the new coordinates of the wireframe vertices. Muscular activities that would produce the estimated deformation are solved through a least squares algorithm. We demonstrate the representative power of muscle force based features on three classifiers; NB, SVM and Adaboost Ability to extract muscle forces that compose a facial expression will enable detection of subtle expressions, replicating an expression on animated characters and exploration of psychologically unknown mechanisms of facial expressions.Yayın An observation based muscle model for simulation of facial expressions(Elsevier Science BV, 2018-05) Erkoç, Tuğba; Ağdoğan, Didem; Eskil, Mustafa TanerThis study presents a novel facial muscle model for coding of facial expressions. We derive this model from unintrusive observation of human subjects in the progress of the surprise expression. We use a generic and single-layered face model which embeds major muscles of the human face. This model is customized onto the human subject's face on the first frame of the video. The last frame of the video is used to project a set of manually marked feature points to estimate the 3 dimensional displacements of vertices due to facial expression. Vertex displacements are used in a mass spring model to estimate the external forces, i.e. the muscle forces on the skin. We observed that the distribution of muscle forces resemble sigmoid or hyperbolic tangent functions. We chose hyperbolic tangent function as our base model and parameterized it using least squares. We compared the proposed muscle model with frequently used models in the literature.Yayın Closeness and uncertainty aware adversarial examples detection in adversarial machine learning(Elsevier Ltd, 2022-07) Tuna, Ömer Faruk; Çatak, Ferhat Özgür; Eskil, Mustafa TanerWhile deep learning models are thought to be resistant to random perturbations, it has been demonstrated that these architectures are vulnerable to deliberately crafted perturbations, albeit being quasi-imperceptible. These vulnerabilities make it challenging to deploy Deep Neural Network (DNN) models in security-critical areas. Recently, many research studies have been conducted to develop defense techniques enabling more robust models. In this paper, we target detecting adversarial samples by differentiating them from their clean equivalents. We investigate various metrics for detecting adversarial samples. We first leverage moment-based predictive uncertainty estimates of DNN classifiers derived through Monte-Carlo (MC) Dropout Sampling. We also introduce a new method that operates in the subspace of deep features obtained by the model. We verified the effectiveness of our approach on different datasets. Our experiments show that these approaches complement each other, and combined usage of all metrics yields 99 % ROC-AUC adversarial detection score for well-known attack algorithms.Yayın Improved microphone array design with statistical speaker verification(Elsevier Ltd, 2021-04) Demir, Kadir Erdem; Eskil, Mustafa TanerConventional microphone array implementations aim to lock onto a source with given location and if required, tracking it. It is a challenge to identify the intended source when the location of the source is unknown and interference exists in the same environment. In this study we combine speaker verification and microphone array processing techniques to localize and maximize gain on the intended speaker under the assumption of open acoustic field. We exploit the steering capability of the microphone array for more accurate speaker verification. Our first contribution is a new N-Gram based and computationally efficient feature for detecting an intended speaker. When the source and interference are localized, microphone array can be tuned further to reduce noise and increase the gain. Our second contribution is this integrated algorithm for speaker verification and localization. In the context of this study we developed SharpEar, an open source environment that simulates propagation of sound emanating from multiple sources. Our third and last contribution is this simulation environment, which is open source and available to researchers of the field.Yayın TENET: a new hybrid network architecture for adversarial defense(Springer Science and Business Media Deutschland GmbH, 2023-08) Tuna, Ömer Faruk; Çatak, Ferhat Özgür; Eskil, Mustafa TanerDeep neural network (DNN) models are widely renowned for their resistance to random perturbations. However, researchers have found out that these models are indeed extremely vulnerable to deliberately crafted and seemingly imperceptible perturbations of the input, referred to as adversarial examples. Adversarial attacks have the potential to substantially compromise the security of DNN-powered systems and posing high risks especially in the areas where security is a top priority. Numerous studies have been conducted in recent years to defend against these attacks and to develop more robust architectures resistant to adversarial threats. In this study, we propose a new architecture and enhance a recently proposed technique by which we can restore adversarial samples back to their original class manifold. We leverage the use of several uncertainty metrics obtained from Monte Carlo dropout (MC Dropout) estimates of the model together with the model’s own loss function and combine them with the use of defensive distillation technique to defend against these attacks. We have experimentally evaluated and verified the efficacy of our approach on MNIST (Digit), MNIST (Fashion) and CIFAR10 datasets. In our experiments, we showed that our proposed method reduces the attack’s success rate lower than 5% without compromising clean accuracy.
