Facial expression synthesis has gained significant attention in the image synthesis field. Generative Adversarial Network (GAN) models have recently gained popularity due to the high-quality synthetic images they prod...
详细信息
ISBN:
(纸本)9798350338737;9798350338720
Facial expression synthesis has gained significant attention in the image synthesis field. Generative Adversarial Network (GAN) models have recently gained popularity due to the high-quality synthetic images they produce. However, these models require complex network architectures that can take days to train, even with high-performance graphicsprocessing Units (GPUs). Many efforts have been made to accelerate and compress such models, but little attention has been paid to the resolution of the images. this study aims to assess the impact of input/output spatial resolution on the resources needed for training a facial expression synthesis model, as well as on the quality of the results. Our results indicate that the produced images and videos had similar quality results measured through objective measures for the spatial resolution of 128 x 128, 256 x 256, and 480 x 480. Furthermore, we found that lower-resolution images could significantly reduce the time required to generate new facial expressions without compromising quality, as measured by objective measures.
Tourette Syndrome (TS) is a genetically induced disorder that is believed to be caused by morphological alterations in brain structure, resulting in involuntary movements known as tics. the current clinical standard f...
详细信息
ISBN:
(数字)9781665453851
ISBN:
(纸本)9781665453851
Tourette Syndrome (TS) is a genetically induced disorder that is believed to be caused by morphological alterations in brain structure, resulting in involuntary movements known as tics. the current clinical standard for diagnosing TS is by clinical assessments performed by physicians. Mild stages of TS, however, commonly go underdiagnosed as tics are infrequent or often suppressed. Brain imaging has been suggested to be a reliable tool to detect brain alterations and possible biomarkers that correspond to neurological disorders. In Magnetic Resonance Imaging (MRI), anatomical brain changes can be identified in the scan by variation in texture patterns of certain regions. the main goal of this work is to identify the statistical significance of texture features in specific brain regions to distinguish TS from control subjects. the proposed approach consists of four main steps: (i) image acquisition, where the data is also organized using demographic information;(ii) brain segmentation, where the structural MRI is parcellated into anatomical regions;(iii) registration, where functional MRI is aligned to structural MRI;(iv) obtaining texture features and statistical analysis, where texture features are extracted from the anatomical brain regions. We adopted 68 subjects aged between 6 to 14 years, divided equally into TS and Normal Control groups. We evaluated the texture features in a statistical manner, where our main findings are: (i) After False Discovery Rate (FDR) correction, only one texture feature was significant (p-value < 0.08) in structural MRI;(ii) Following FDR correction, eight texture features in functional MRI for three anatomical regions were considered significant;(iii) the right amygdala presented significance in distinct texture features, matching its importance in the literature. Texture features aligned withthe literature can serve as a reliable tool to identify imaging changes, which can lead to future work applied to clinical studies.
Deep neural networks are extensively used for solving a variety of computer vision problems. However, in order for these networks to obtain good results, a large amount of data is necessary for training. In image clas...
详细信息
ISBN:
(纸本)9781665423540
Deep neural networks are extensively used for solving a variety of computer vision problems. However, in order for these networks to obtain good results, a large amount of data is necessary for training. In image classification, this training data consists of images and labels that indicate the class portrayed by each image. Obtaining this large labeled dataset is very time and resource consuming. therefore, domain adaptation methods allow different, but semantic-related, datasets that are already labeled to be used during training, thus eliminating the labeling cost. In this work, the effects of embedding dimensionality reduction in a state-of-the-art domain adaptation method are analyzed. Furthermore, we experiment with a different approach that use the available data from all domains to compute the confidence of pseudo-labeled samples. We show through experiments in commonly used datasets that, in fact, the proposed modifications led to better results in the target domain in some scenarios.
this paper presents a new technique to solve the single image super resolution reconstruction problem based on multiple extreme learning machine regressors, called here MELM. the MELM employs a feature space of low re...
详细信息
ISBN:
(纸本)9781538622193
this paper presents a new technique to solve the single image super resolution reconstruction problem based on multiple extreme learning machine regressors, called here MELM. the MELM employs a feature space of low resolution images, divided in subspaces, and one regressor is trained for each one. In the training task, we employ a color dataset containing 91 images, with approximately 5.3 million pixels, and PSNR and SSIM as metric evaluation. For the experiments we use two datasets, Set 5 and Set 14, to evaluate the results. We observe MELM improves reconstruction quality in about 0.44 dB PSNR in average for Set 5, when compared with a global ELM regressor (GELM), trained for the entire feature space. the proposed method almost reaches deep learning reconstruction quality, without depending on large datasets and long training times, giving a competitive trade off between performance and computational costs.
the human brain is able to rapidly understand scenes through the recognition of their composing elements and comprehension of the role that each of them plays. this process, related to human perception, impacts in wha...
详细信息
ISBN:
(纸本)9781538622193
the human brain is able to rapidly understand scenes through the recognition of their composing elements and comprehension of the role that each of them plays. this process, related to human perception, impacts in what people care when they see an image and the priority they give to each element. the idea of priority, also referred as importance, is based on biological features of perception and social aspects that interfere in how people perceive what they see and what is considered relevant. In this context, this paper proposes the Element Importance Relative Assignment (EIRA), an approach that models how humans attribute importance to elements in a scene. this approach is based on perceptual, compositional and contextual features employed to assign importance to elements in a scene. To evaluate the proposed approach, tests are conducted in different image datasets with emphasis on the UIUC Pascal Sentence Dataset, where our approach achieves an average accuracy of 86.89%.
We present further results of the application of Dempster-Shafer theory for uncertainty reasoning in corresponding distorted images in computer vision. In a previous work (J.D.S. Silva et al., 2002), the model was app...
详细信息
We present further results of the application of Dempster-Shafer theory for uncertainty reasoning in corresponding distorted images in computer vision. In a previous work (J.D.S. Silva et al., 2002), the model was applied to correspond radiometrically distorted images, that is, images with differences in brightness and contrast, as an extension of the work developed by J.D.S. Silva et al. (2001). the results showed the model is robust when dealing with pairs of nonequalized images and encouraged us to try to correspond geometric distorted images, that is, pairs of images in which one is rotated in relation to the other. In the conducted experiments, the right image was rotated by different angles to simulate the desired geometric distortions. the model was applied to a pair of rotated images and it successfully established the correspondence of a pair of points. As in previous works, the correspondence evidences are based on the contextual and structural features of the points, and their combination is performed by Dempster-Shafer's rule of combination for uncertainty reasoning. A search process maximizes the belief on the combined evidences.
In several domains a refinement criterion is often needed to decide whether to go on or to stop sampling a signal. When the sampled values are homogeneous enough, we assume that they represent the signal fairly well a...
详细信息
Flow visualization has been an active research field for several years and streamlines have proved to be an effective representation of two-dimensional steady vector fields. Online simulation of complex dynamic system...
详细信息
Flow visualization has been an active research field for several years and streamlines have proved to be an effective representation of two-dimensional steady vector fields. Online simulation of complex dynamic systems, computed on remote powerful computers are becoming more common. Efficient steering of the computational process requires the analysis of simulation results in real time. Visualization offers the possibility of at-a-glance analysis of the data but requires being able to download visualizations efficiently. Here we propose a wavelet model and compression scheme for streamline-based images of 2D steady vector fields. this method allows us to decrease the size of the data by a factor of 2.5 for a visualization quality comparable to the original image, and the compression factor increases to 10 if we accept small deformations. (C) 2002 Elsevier Science Ltd. All rights reserved.
the proceedings contain 85 papers. the topics discussed include: ongoing learning for supervised pattern recognition;performance evaluation of prototype selection algorithms for nearest neighbor classification;on reco...
ISBN:
(纸本)0769513301
the proceedings contain 85 papers. the topics discussed include: ongoing learning for supervised pattern recognition;performance evaluation of prototype selection algorithms for nearest neighbor classification;on reconstructing surfaces of arbitrary topology from a range image;multi-resolution classification trees in OCR design;an OOP approach for mesh generation of multi-region models with NURBS;high-level verification of handwritten numeral strings;discrete scale spaces via heat equation;hybrid human-machine non-linear filter design using envelopes;procedural models on image synthesis for ocean animation;fast multidimensional parallel Euclidean distance transform based on mathematical morphology;connected filtering by gray-level classification through morphological histogram processing;multi-bands image analysis using local fractal dimension;microarray gridding by mathematical morphology;procedural shape synthesis on subdivision surfaces;and dynamic algorithm binding for interactive walkthroughs.
A tool created withthe aim of contributing to the teaching of imageprocessing techniques is presented. It allows the specification of filters in a simple and intuitive manner. Being a Java application, the system is...
详细信息
ISBN:
(纸本)0769513301
A tool created withthe aim of contributing to the teaching of imageprocessing techniques is presented. It allows the specification of filters in a simple and intuitive manner. Being a Java application, the system is portable and runs in many different environments.
暂无评论