convolutional neural network

Fault detection in photovoltaic (PV) arrays is necessary to increase the output power and also the useful life of a PV system. The presence of conditions such as partial shade, high impedance faults, and the maximum power point detector (MPPT) system make the fault detection of PV in environmental conditions more challenging. The literature identified and classified defects just in few scenarios. In this study two-dimensional scalograms are generated from PV system data. The VGG16 as a pretrained convolutional neural network is used for feature extraction. Finally, to identify and classify faults in the PV system a fully connected neural network is trained. Unlike the previous methods proposed in the literature on the subject of defect detection and classification, various defective cases with MPPT combination are considered in this research. It has been shown that the proposed method including pre-trained CNN performs better than the existing methods and achieves an error detection accuracy of 83.375%.

Medical image segmentation is one of the most important steps in medical image analysis to improve diagnosis and findings. One of the most common segmentation methods in deep learning is the use of Unet networks. The presence of overlapping layers in the Unet decoder does not allow extracting information from deeper layers. Also, due to the limited range of the received field of convolution cores, long-range information and dependencies are not considered well. In this article, our goal is to place a structure in the area between encoder and decoder in the Unet model in order to fill the semantic gap between the encoder and decoder area and better extract features by paying attention to local and global features. This model makes the target region more prominent in different medical datasets. We have conducted our experiment on 6 medical data sets, and the results obtained in two evaluation criteria, Dice and Iou, show that our proposed model has better results than Unet and based methods.

Steganalysis is the art of detecting the existence of hidden data. Recent research has revealed that convolutional neural networks (CNNs) can detect data through automatic feature extraction. Several studies investigated the performance of existing models using a limited number of spatial steganography methods. This study aims to propose a CNN and comprehensively investigate its efficiency in detecting different spatial methods. The proposed model comprises three modules: preprocessing, convolutional (five blocks), and classifier (three fully connected layers). The test results for the least-significant-bit (LSB) and pixel-value differencing (PVD) based methods indicate that the proposed method can detect data of even concise length with high accuracy and a low error. The proposed method also detects complexity-based LSB-M (CBL) as an adaptive approach. Lower embedding rates make this success even more impressive. Manual feature extraction has much lower success rates due to low variations of statistical features at low embedding rates than the proposed model.

Image captioning is a new field of artificial intelligence which uses both natural language processing and machine learning methods. Recently, several researches conducted on Image captioning have led to the acceptable results in this field. Moreover, the need of the medical community for automatic interpretation of medical images, prompted researchers to apply this approach in the field of automatic interpretation of medical images. The medical images captioning is more challenging than the problem of image description because medical images are mostly gray and accompanied by noise, the existing data sets have significantly few samples; On the other hand, the provided interpretations are not structured and have variable lengths. In order to achieve the desired results, many algorithms have been proposed, among which methods based on deep learning have achieved proper results. This article introduces datasets, evaluation criteria and methods developed based on deep learning in the field of automatic interpretation of medical images in order to help understand the existing literature and highlight future directions in this field.

Automatic image captioning is a challenging task in computer vision and aims to generate computer-understandable descriptions for images. Employing convolutional neural networks (CNN) has a key role in image caption generation. However, during the process of generating descriptions for an image, there are two major challenges for CNN, such as: they do not consider the relationships and spatial hierarchical structures between the objects in the image, and the lack of resistance against rotational changes of the images. In order to solve these challenges, this paper presents an improved capsule network to describe image content using natural language processing by considering the relations between the objects . A capsule contains a set of neurons that consider the parameters of the state of objects in the image, such as size, direction, scale, and relationships of objects to each other. These capsules have a special focus on extracting meaningful features for use in the process of generating relevant descriptions for a given set of images. Qualitative tests on the MS-COCO dataset using the capsule network and ELMo embedding technique have resulted in 2-5% improvement in the evaluated metrics compared to existing image captioning models.

The welcoming of social networks, especially Twitter, has posed a new challenge to researchers, and it is nothing but spam. Numerous different approaches to deal with spam are presented. In this study, we attempt to enhance the accuracy of spam detection by applying one of the latest spam detection techniques and its combination with sentiment analysis. Using the word embedding technique, we give the tweet text as input to a convolutional neural network (CNN) architecture, and the output will detect spam text or normal text. Simultaneously, by extracting the suitable features in the Twitter network and applying machine learning methods to them, we separately calculate the Tweeter spam detection. Eventually, we enter the output of both approaches into a Meta Classifier so that its output specifies the final spam detection or the normality of the tweet text. In this study, we employ both balanced and unbalanced datasets to examine the impact of the proposed model on two types of data. The results indicate an increase in the accuracy of the proposed method in both datasets.

Coronavirus 2019, or Covid-19, is an acute respiratory disease with high virus transmission capacity that has led to high mortality rates worldwide. Although rapid diagnosis can play an essential role in the patient's recovery, radiography by the treatment staff is a time-consuming process.Therefore, the use of ultrasound data and deep learning techniques are recommended. The ultrasound technique is radiation-free and can be used in pediatric wards and intensive care units for specific patients. However, its data have noise that affects the performance of deep learning methods. To this end, in this paper, we combine the uniform local noise-resistant binary pattern method with the deep learning method. First, a uniform local binary pattern is calculated on two temporal planes to extract the features of the Covid-19 manifestations in consecutive ultrasound data, and then, the resulting matrix is given as the input of the convolutional network. According to the performed experiments, the proposed method has better performance compared to that of the other state-of-art methods. The results show that the diagnosis accuracy of the Covid-19 from the ultrasound data using the proposed method is 98.5%.

Due to the significant growth of textual information and data generated by humans on social networks, there is a need for systems that can automatically analyze the data and extract valuable information from them. One of the most important textual data is people's opinions about a particular topic that are expressed in the form of text. Text published by users on social networks can represent their personality. Although machine learning based methods can be considered as a good choice for analyzing these data, there is also a remarkable need for deep learning based methods to overcome the complexity and dispersion of content and syntax of textual data during the training process. In this regard, the purpose of this paper is to employ deep learning based methods for personality recognition. Accordingly, the convolutional neural network is combined with the Adaboost algorithm to consider the possibility of using the contribution of various filter lengths and gasp their potential in the final classification via combining various classifiers with respective filter sizes using AdaBoost. The proposed model was conducted on Essays and YouTube datasets. Based on the empirical results, the proposed model presented superior performance compared to other existing models on both datasets.