Deep neural networks are one of the most important branches of machinelearning that have been recently used in many fields of patternrecognition and machine vision applications successfully. One of the most famous n...
详细信息
ISBN:
(纸本)9781538695692
Deep neural networks are one of the most important branches of machinelearning that have been recently used in many fields of patternrecognition and machine vision applications successfully. One of the most famous networks in this area is convolutional neural networks which are biologically inspired variants of multi-layer perceptions. In these networks, activation function plays a significant role especially when the data come in different scales. Recently, there is an interest to adaptive activation functions which adapts their parameters to the input data during network training process. Therefore, in this paper, inspired from a successful convolutional neural network tuned for medical image classification, we have investigated the effect of applying adaptive activation functions in a modified convolutional network by combining basic activation functions in linear (mixed) and nonlinear (gated) ways. The effectiveness of using these adaptive functions is shown on a CT brain images dataset (as a complex medical dataset) and the well-known MN 1st hand-written digits dataset. The done experiments show that the classification accuracy of the proposed network with adaptive activation functions is higher compared to the ones using basic activation functions.
Multiple instance learning (MIL) is a form of weakly supervised learning for problems in which training instances are arranged into bags, and a label is provided for whole bags but not for individual instances. Most p...
详细信息
ISBN:
(纸本)9781467389105
Multiple instance learning (MIL) is a form of weakly supervised learning for problems in which training instances are arranged into bags, and a label is provided for whole bags but not for individual instances. Most proposed MIL algorithms focus on bag classification, but more recently, the classification of individual instances has attracted the attention of the patternrecognition community. While these two tasks are similar, there are important differences in the consequences of instance misclassification. In this paper, the scoring function learned by MIL classifiers for the bag classification task is exploited for instance classification by adjusting the decision threshold. A new criterion for the threshold adjustment is proposed and validated using 7 reference MIL algorithms on 3 real-world data sets from different application domains. Experiments show considerable improvements in accuracy over these algorithms for instance classification. In some applications, the unweighted average recall increases by as much as 18%, while the F-1-score increases by 12%.
In this paper we present a survey of current research in Music Information Retrieval in North Indian Classical Music and describe all the characteristics of ragas used for classification. We then describe Bhatkhande...
详细信息
ISBN:
(纸本)9789811020353;9789811020346
In this paper we present a survey of current research in Music Information Retrieval in North Indian Classical Music and describe all the characteristics of ragas used for classification. We then describe Bhatkhande's classification scheme and show how it can simplify the classification process of 100 ragas to 10 categories. We also discuss the issues that need to be addressed and the similarities and differences between Hindustani classical music and Western Classical music. Current research efforts on Raga identification are also described.
The segmentation of the surgical workflow might be helpful for providing context-sensitive user interfaces, or generating automatic report. Our approach focused on the automatic recognition of surgical phases by micro...
详细信息
ISBN:
(纸本)9783642137105
The segmentation of the surgical workflow might be helpful for providing context-sensitive user interfaces, or generating automatic report. Our approach focused on the automatic recognition of surgical phases by microscope image classification. Our workflow, including images features extraction, image database labelisation, Principal Component Analysis (PCA) transformation and 10-fold cross-validation studies was performed on a specific type of neurosurgical intervention, the pituitary surgery. Six phases were defined by an expert for this type of intervention. We thus assessed machinelearning algorithms along with the data dimension reduction. We finally kept 40 features from the PCA and found a best correct classification rate of the surgical phases of 82% with the multiclass Support Vector machine.
In this work, we processed sets of images obtained by the light-sheet fluorescence microscopy method. We selected different cell groups and determined areas occupied by ensembles of cell groups in mouse brain tissue. ...
详细信息
ISBN:
(数字)9783030954673
ISBN:
(纸本)9783030954673;9783030954666
In this work, we processed sets of images obtained by the light-sheet fluorescence microscopy method. We selected different cell groups and determined areas occupied by ensembles of cell groups in mouse brain tissue. recognition of mouse neuronal populations was performed on the basis of visual properties of fluorescence-activated cells. Individual elements were selected based on their brightness in grayscale mode. Methods of spatial data processing were applied to identify border areas between ensembles and to calculate topological characteristics of cell groups. By applying cell statistics operations, we obtained the localization of the regions of interest, for subsequent identification of samples with specified topological characteristics. Based on the topological properties of the cell groups, we constructed training samples, and then used these to detect typical sets of ensembles in multi-page TIFF files with optogenetics datasets.
Although digitization is advancing rapidly, a large amount of data processed by companies is in printed format. Technologies such as Optical Character recognition (OCR) support the transformation of printed text into ...
详细信息
ISBN:
(纸本)9783031065156;9783031065163
Although digitization is advancing rapidly, a large amount of data processed by companies is in printed format. Technologies such as Optical Character recognition (OCR) support the transformation of printed text into machine-readable content. However, OCR struggles when data on documents is highly unstructured and includes non-text objects. This, e.g., applies to documents such as medical prescriptions. Leveraging Design Science Research (DSR), we propose a flexible processing pipeline that can deal with character recognition on the one hand and object detection on the other hand. To do so, we derive Design Requirements (DR) in cooperation with a practitioner doing prescription billing in the healthcare domain. We then developed a prototype blueprint that is applicable to similar problem formulations. Overall, we contribute to research and practice in multiple ways. First, we provide evidence for selected OCR methods provided by previous research. Second, we design a machine-learning-based digitization pipeline for printed documents containing both text and non-text objects in the context of medical prescriptions. Third, we derive a nascent design pattern for this type of document digitization. These patterns are the foundation for further research and can support the development of innovative information systems leading to more efficient decision making and thus to economic resource usage.
In this paper, a new image analysis method using region and formula based approach is proposed for automated illicit material detection. The main purpose of this method is to reveal the true gray levels of the object ...
详细信息
In this paper, a new image analysis method using region and formula based approach is proposed for automated illicit material detection. The main purpose of this method is to reveal the true gray levels of the object of interest by removing the background object effects using the neighboring region's information. The preliminary results have shown great promising of this method.
The use of mean field decomposition in Markov random field (MRF) based unsupervised textured image segmentation was discussed. The method was based on the expectation and maximization (EM) method where a decomposed po...
详细信息
The use of mean field decomposition in Markov random field (MRF) based unsupervised textured image segmentation was discussed. The method was based on the expectation and maximization (EM) method where a decomposed posteriori probability was used. A posteriori probability of the whole region image was decomposed into the product of local a posteriori probabilities (LAP) for all pixels. It was found that the use of LAP was essential to perform better image segmentation.
This paper presents a general methodology for the development of fuzzy algorithms for learning vector quantization (FALVQ). These algorithms can be used to train feature maps to perform pattern clustering through an u...
详细信息
ISBN:
(纸本)0819418455
This paper presents a general methodology for the development of fuzzy algorithms for learning vector quantization (FALVQ). These algorithms can be used to train feature maps to perform pattern clustering through an unsupervised learning process. The development of FALVQ algorithms is based on the minimization of a fuzzy objective function, formed as the weighted sum of the squared Euclidean distances between an input vector, which represents a feature vector, and the weight vectors of the map, which represent the prototypes. This formulation leads to the development of genuinely competitive algorithms, which allow all prototypes to compete for matching each input. The FALVQ 1, FALVQ 2, and FALVQ 3 families of algorithms are developed by selecting admissible generalized membership functions with different properties. The efficiency of the proposed algorithms is illustrated by their use in codebook design required for image compression based on vector quantization.
Oil exploration mainly targets to the locations that are closed or below the salt bodies, in the underlying geologic structure. With time the computational tools which can help in interpreting, analysing and estimatin...
详细信息
ISBN:
(纸本)9781467368094
Oil exploration mainly targets to the locations that are closed or below the salt bodies, in the underlying geologic structure. With time the computational tools which can help in interpreting, analysing and estimating the geometry with its position has been increased. But still at many time the data which is gathered using these computational tools is recognized with the lack of resolution and poor structural identification which create severe technical and economic problems. Under such circumstances, seismic interpretation based only on the humaneye is inaccurate. In such a situation, for making good decisions and production planning all things depend on good- quality seismic images that generally are not feasible in salt tectonics areas. Therefore, a generalization of the Hough transform is applied to build parabolic, line, circular and arbitrary shapes that are useful in the idealization and recognition of salt domes from 2D seismic profiles. The contribution of this project is oriented in providing the seismic interpreters with semi-automatic computational tool. Hence, the paper focuses on imageprocessing technique namely Hough transform along with sobel as well as canny algorithm which are applied to detect and delineate complex salt bodies from seismic exploration profiles. The novel imageprocessing approach presented here will be helpful in the identification of complex geological features from seismic images.
暂无评论