8 sonuçlar
Arama Sonuçları
Listeleniyor 1 - 8 / 8
Yayın Anatomy based animation of facial expressions(Işık Üniversitesi, 2012-12-24) Erkoç, Tuğba; Eskil, Mustafa Taner; Işık Üniversitesi, Fen Bilimleri Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans ProgramıThis study presents a new physics based facial expression animation system based on anatomic data. Proposed system consists a face model and a new facial expression animation generation algorithm. The proposed system is a Mass-Spring-Damper (MSD) system. It consists single layer face model HIGEM and a set of facial muscles which are placed anatomically correct places. Non-linear viscoelastic characteristics of the human skin is approximated with non-linear springs. The set of muscles triggers facial expressions. HIGEM does not include a skull to support the facial mesh, so it tends to collapse under muscle forces. A new algorithm proposed to prevent this. This algorithm uses back and forward projections for re-defining the muscle forces. Another problem of MSD systems is individual element collapse under large forces. This is addressed with Edge Repulsion (ER) approach. Dynamics of the face is modelled with a quasi-implicit Ordinary Differential Equation (ODE). The regions of the face affected by the facial muscles were computed and marked offline in previous works to speed up the animations. It requires offline re-calculation whenever the activation level of the muscles or facial mesh topology changes. Thus, a new generic stack based approach which works at runtime is proposed. The proposed system is tested with eight facial expressions; happy, sad, angry, feared, surprised, disgusted, disgusted anger and happy surprise. Elapsed times for the facial expressions vary with number of contracting muscles, their region of influence, and the time value for each timestep. Keywords: Facial expression animation, physics based approach, generic wireframe, mass-spring-damper system.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 A novel similarity based unsupervised technique for training convolutional filters(IEEE, 2023-05-17) Erkoç, Tuğba; Eskil, Mustata TanerAchieving satisfactory results with Convolutional Neural Networks (CNNs) depends on how effectively the filters are trained. Conventionally, an appropriate number of filters is carefully selected, the filters are initialized with a proper initialization method and trained with backpropagation over several epochs. This training scheme requires a large labeled dataset, which is costly and time-consuming to obtain. In this study, we propose an unsupervised approach that extracts convolutional filters from a given dataset in a self-organized manner by processing the training set only once without using backpropagation training. The proposed method allows for the extraction of filters from a given dataset in the absence of labels. In contrast to previous studies, we no longer need to select the best number of filters and a suitable filter weight initialization scheme. Applying this method to the MNIST, EMNIST-Digits, Kuzushiji-MNIST, and Fashion-MNIST datasets yields high test performances of 99.19%, 99.39%, 95.03%, and 90.11%, respectively, without applying backpropagation training or using any preprocessed and augmented data.Yayın Subset selection for tuning of hyper-parameters in artificial neural networks(IEEE, 2017) Aki, K.K.Emre; Erkoç, Tuğba; Eskil, Mustafa TanerHyper-parameters of a machine learning architecture define its design. Tuning of hyper-parameters is costly and for large data sets outright impractical, whether it is performed manually or algorithmically. In this study we propose a Neocognitron based method for reducing the training set to a fraction, while keeping the dynamics and complexity of the domain. Our approach does not require processing of the entire training set, making it feasible for larger data sets. In our experiments we could successfully reduce the MNIST training data set to less than 2.5% (1,489 images) by processing less than 10% of the 60K images. We showed that the reduced data set can be used for tuning of number of hidden neurons in a multi-layer perceptron.Yayın Object recognition with competitive convolutional neural networks(Işık Üniversitesi, 2023-06-12) Erkoç, Tuğba; Eskil, M. Taner; Işık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Doktora Programı; Işık University, School of Graduate Studies, Ph.D. in Computer EngineeringIn recent years, Artificial Intelligence (AI) has achieved impressive results, often surpassing human capabilities in tasks involving language comprehension and visual recognition. Among these, computer vision has experienced remarkable progress, largely due to the introduction of Convolutional Neural Networks (CNNs). CNNs are inspired by the hierarchical structure of the visual cortex and are designed to detect patterns, objects, and complex relationships within visual data. One key advantage is their ability to learn directly from pixel values without the need for domain expertise, which has contributed to their popularity. These networks are trained using supervised backpropagation, a process that calculates gradients of the network’s parameters (weights and biases) with respect to the loss function. While backpropagation enables impressive performance with CNNs, it also presents certain drawbacks. One such drawback is the requirement for large amounts of labeled data. When the available data samples are limited, the gradients estimated from this limited information may not accurately capture the overall data behavior, leading to suboptimal parameter updates. However, obtaining a sufficient quantity of labeled data poses a challenge. Another drawback is the requirement of careful configuration of hyperparameters, including the number of neurons, learning rate, and network architecture. Finding optimal values for these hyperparameters can be a time-consuming process. Furthermore, as the complexity of the task increases, the network architecture becomes deeper and more complex. To effectively train the shallow layers of the network, one must increase the number of epochs and experiment with solutions to prevent vanishing gradients. Complex problems often require a greater number of epochs to learn the intricate patterns and features present in the data. It’s important to note that while CNNs aim to mimic the structure of the visual cortex, the brain’s learning mechanism does not necessarily involve back-propagation. Although CNNs incorporate the layered architecture of the visual cortex, the reliance on backpropagation introduces an artificial learning procedure that may not align with the brain’s actual learning process. Therefore, it is crucial to explore alternative learning paradigms that do not rely on backpropagation. In this dissertation study, a unique approach to unsupervised training for CNNs is explored, setting it apart from previous research. Unlike other unsupervised methods, the proposed approach eliminates the reliance on backpropagation for training the filters. Instead, we introduce a filter extraction algorithm capable of extracting dataset features by processing images only once, without requiring data labels or backward error updates. This approach operates on individual convolutional layers, gradually constructing them by discovering filters. To evaluate the effectiveness of this backpropagation-free algorithm, we design four distinct CNN architectures and conduct experiments. The results demonstrate the promising performance of training without backpropagation, achieving impressive classification accuracies on different datasets. Notably, these outcomes are attained using a single network setup without any data augmentation. Additionally, our study reveals that the proposed algorithm eliminates the need to predefine the number of filters per convolutional layer, as the algorithm automatically determines this value. Furthermore, we demonstrate that filter initialization from a random distribution is unnecessary when backpropagation is not employed during training.Yayın TurkEmbed: Turkish embedding model on natural language inference & sentence text similarity tasks(Institute of Electrical and Electronics Engineers Inc., 2025) Ezerceli, Özay; Gümüşçekiçci, Gizem; Erkoç, Tuğba; Özenç, BerkeThis paper introduces TurkEmbed, a novel Turkish language embedding model designed to outperform existing models, particularly in Natural Language Inference (NLI) and Semantic Textual Similarity (STS) tasks. Current Turkish embedding models often rely on machine-translated datasets, potentially limiting their accuracy and semantic understanding. TurkEmbed utilizes a combination of diverse datasets and advanced training techniques, including matryoshka representation learning, to achieve more robust and accurate embeddings. This approach enables the model to adapt to various resource-constrained environments, offering faster encoding capabilities. Our evaluation on the Turkish STS-b-TR dataset, using Pearson and Spearman correlation metrics, demonstrates significant improvements in semantic similarity tasks. Furthermore, TurkEmbed surpasses the current state-of-the-art model, Emrecan, on All-NLI-TR and STS-b-TR benchmarks, achieving a 1-4% improvement. TurkEmbed promises to enhance the Turkish NLP ecosystem by providing a more nuanced understanding of language and facilitating advancements in downstream applications.Yayın TurkEmbed4Retrieval: Türkçe için geri getirme görevine özel gömme modeli(Institute of Electrical and Electronics Engineers Inc., 2025-08-15) Ezerceli, Özay; Gümüşçekiçci, Gizem; Erkoç, Tuğba; Özenç, BerkeBu çalışmada, öncelikle Doğal Dil Çıkarımı (DDÇ) ve Anlamsal Metin Benzerliği (AMB) görevleri için geliştirilen TurkEmbed modelinin, MS-Marco-TR veri seti üzerinde ince ayar yapılarak geri getirme görevlerine uygun hale getirilmesini sağlayan TurkEmbed4Retrieval modelini tanıtıyoruz. Model, Matruşka temsili ögrenme ve özel tasarlanmış negatif çiftlerin sıralanması kayıp fonksiyonu gibi ileri seviye egitim teknikleri kullanılarak optimize edilmiştir. Yapılan kapsamlı deneyler, TurkEmbed4Retrieval’ın, geri getirme metriklerinde TurkishcolBERT modelini Scifact-TR veri kümesinde %19–26 oranında geçtiğini göstermektedir. Bu bağlamda, modelimiz, Türkçe bilgi getirme sistemleri için yeni bir çıtaya ulaşmaktadır.Yayın El yazısı rakam sınıflandırması için gözetimsiz benzerlik tabanlı evrişimler(Institute of Electrical and Electronics Engineers Inc., 2022) Erkoç, Tuğba; Eskil, Mustafa TanerEffective training of filters in Convolutional Neural Networks (CNN) ensures their success. In order to achieve good classification results in CNNs, filters must be carefully initialized, trained and fine-tuned. We propose an unsupervised method that allows the discovery of filters from the given dataset in a single epoch without specifying the number of filters hyper-parameter in convolutional layers. Our proposed method gradually builds the convolutional layers by a discovery routine that extracts a number of features that adequately represent the complexity of the input domain. The discovered filters represent the patterns in the domain, so they do not require any initialization method or backpropagation training for fine tuning purposes. Our method achieves 99.03% accuracy on MNIST dataset without applying any data augmentation techniques.
