×
The submission system is temporarily under maintenance. Please send your manuscripts to
Go to Editorial ManagerIn recent times, artificial intelligence has become an essential part of our lives, particularly in tasks involving object recognition. This paper explores the use of convolutional neural networks (CNNs) for enhancing underwater search and rescue operations by classifying images of humans, fish, and plants. Leveraging the OpenCV library for preprocessing and the Keras library with a TensorFlow backend for recognition, this study utilizes a dataset captured through field experiments. The methodology involved preprocessing the images for segmentation, followed by training a CNN model to classify these images with high accuracy. The CNN model demonstrated a remarkable classification accuracy of 99.6 %, significantly outperforming other modern machine-learning methods. This work suggests that CNNs can greatly improve the speed and effectiveness of underwater search and rescue operations by accurately identifying and locating submerged persons, which is critical for timely rescue missions.
Accurate long-term load forecasting (LTLF) is crucial for smart grid operations, but existing CNN-based methods face challenges in extracting essential featuresfrom electricity load data, resulting in diminished forecasting performance. To overcome this limitation, we propose a novel ensemble model that integratesa feature extraction module, densely connected residual block (DCRB), longshort-term memory layer (LSTM), and ensemble thinking. The feature extraction module captures the randomness and trends in climate data, enhancing the accuracy of load data analysis. Leveraging the DCRB, our model demonstrates superior performance by extracting features from multi-scale input data, surpassing conventional CNN-based models. We evaluate our model using hourly load data from Odisha and day-wise data from Delhi, and the experimental results exhibit low root mean square error (RMSE) values of 0.952 and 0.864 for Odisha and Delhi, respectively. This research contributes to a comparative long-term electricity forecasting analysis, showcasing the efficiency of our proposed model in power system management. Moreover, the model holds the potential to sup-port decisionmaking processes, making it a valuable tool for stakeholders in the electricity sector.
Federated learning (FL) is one of the newest and most significant fields for developing artificial intelligence applications. This technology trains its models in a distributed way, using data from different clients who work together in the system without sharing their data. The training process is kept local to protect the privacy of the data. Among the many difficulties that have arisen due to the novelty of this technology is the issue of heterogeneous data between typical clients. Client’s data may differ from each other in different respects, for example non identically and independent distribution (non-IID) between clients and the difference in the type of data used in each client. This can lead to inconsistencies in the model’s predictions and other undesirable outcomes. This paper discussed ways to solve this problem where clients with heterogeneous data were dealt with in terms of number and type. Because there are different types of image data through which doctors can diagnose coronavirus, such as x-rays, CT-scan. A hybrid convolution neural network (CNN ) and long short-term memory model (LSTM) has been proposed in a federated learning system to predict the incidence of this disease by using two clients, each with one of these different data. Good results were obtained with an accuracy of more than 99% in one customer and more than 95% in the second client while maintaining the privacy of this data.
Recently, face recognition technology has become more prevalent in various applications, including mobile devices, access control, and financial transactions. Therefore, it is crucial to address potential vulnerabilities that attackers might exploit. In this study, a method for face presentation attack detection (PAD) is introduced. The method utilizes the diversity of modalities provided by some cameras and sensors to detect face spoofing using convolutional neural networks (CNN) within the context of deep learning. To assess the effectiveness of the proposed approach in real-world scenarios, the wide multi-channel presentation attack (WMCA) dataset is used. The presented method exploits the multi-modal data, including RGB, depth, IR, and thermal channels, to enhance system performance and explore different techniques for combining the results from each modality. Furthermore, this study explores diverse techniques for fusing results from each channel in two fusion scenarios, pre-fusion and post-fusion. In the pre-fusion scenario, data from the four channels is combined, resulting in an ACER value of 0.19%. In the post-fusion scenario, the results of each modality are fused using different fusion techniques, such as majority voting, weighted voting, average pooling, and a stacking classifier. The stacking classifier yields the most favorable outcome with an ACER ratio of 0.03%. This performance is notably superior when compared to state-of-the-art methodologies.
In recent years, there has been a considerable rise in the applications in which object or image categorization is beneficial for example, analyzing medicinal images, assisting persons to organize their collections of photos, recognizing what is around self-driving vehicles, and many more. These applications necessitate accurately labeled datasets, in their majority involve an extensive diversity in the types of images, from cats or dogs to roads, landscapes, and so forth. The fundamental aim of image categorization is to predict the category or class for the input image by specifying to which it belongs. For human beings, this is not a considerable thing, however, learning computers to perceive represents a hard issue that has become a broad area of research interest, and both computer vision techniques and deep learning algorithms have evolved. Conventional techniques utilize local descriptors for finding likeness between images, however, nowadays; progress in technology has provided the utilization of deep learning algorithms, especially the Convolutional Neural Networks (CNNs) to auto-extract representative image patterns and features for classification The fundamental aim of this paper is to inspect and explain how to utilize the algorithms and technologies of deep learning to accurately classify a dataset of images into their respective categories and keep model structure complication to a minimum. To achieve this aim, must focus precisely and accurately on categorizing the objects or images into their respective categories with excellent results. And, specify the best deep learning-based models in image processing and categorization. The developed CNN-based models have been proposed and a lot of pre-training models such as (VGG19, DenseNet201, ResNet152V2, MobileNetV2, and InceptionV3) have been presented, and all these models are trained on the Caltech-101 and Caltech-256 datasets. Extensive and comparative experiments were conducted on this dataset, and the obtained results demonstrate the effectiveness of the proposed models. The obtained results demonstrate the effectiveness of the proposed models. The accuracy for Caltech-101 and Caltech-256 datasets was (98.06% and 90%) respectively.
Self-driving cars are a fundamental research subject in recent years; the ultimate goal is to completely exchange the human driver with automated systems. On the other hand, deep learning techniques have revealed performance and effectiveness in several areas. The strength of self-driving cars has been deeply investigated in many areas including object detection, localization as well, and activity recognition. This paper provides an approach to deep learning; which combines the benefits of both convolutional neural network CNN together with Dense technique. This approach learns based on features extracted from the feature extraction technique which is linear discriminant analysis LDA combined with feature expansion techniques namely: standard deviation, min, max, mod, variance and mean. The presented approach has proven its success in both testing and training data and achieving 100% accuracy in both terms.
License plate recognition is an essential part of contemporary surveillance systems since it is helpful in many applications, including parking management, vehicle access control, traffic control, and law enforcement. This project aims to provide a robust and dependable method for detecting license plates that will outperform existing approaches in accuracy and dependability. This observation method uses contemporary technology to address challenging troubles related to license plate recognition. Our methodology is primarily based on the Faster R-CNN structure, an established model for picture item detection. The novel thing, even though, is how Gray Wolf Optimization—which draws notion from the searching conduct of gray wolves—is mixed with the Faster R-CNN network. The accuracy is greatly improved by this synergistic combination, which also improves detection abilities. Moreover, an improved ResNet-50 model is blanketed to improve the classification system similarly, ensuring accurate license plate detection in several situations. The extensively utilized ”car license plate detection” dataset is used to assess the recommended technology very well, confirming its efficacy in practical settings. The empirical outcomes show exceptional performance, with a median precision of 98.21%, demonstrating how nicely the hybrid method works to attain the very best stage of license plate detecting accuracy. This painting establishes a new benchmark in license plate identity using cutting-edge technology and innovative techniques, starting the door for enhanced safety and surveillance.
Arial images are very high resolution. The automation for map generation and semantic segmentation of aerial images are challenging problems in semantic segmentation. The semantic segmentation process does not give us precise details of the remote sensing images due to the low resolution of the aerial images. Hence, we propose an algorithm U-Net Architecture to solve this problem. It is classified into two paths. The compression path (also called: the encoder) is the first path and is used to capture the image's context. The encoder is just a convolutional and maximal pooling layer stack. The symmetric expanding path (also called: the decoder) is the second path, which is used to enable exact localization by transposed convolutions. This task is commonly referred to as dense prediction, which is completely connected to each other and also with the former neurons which gives rise to dense layers. Thus it is an end-to-end fully convolutional network (FCN), i.e. it only contains convolutional layers and does not contain any dense layer because of which it can accept images of any size. The performance of the model will be evaluated by improving the image using the proposed method U-NET and obtaining an improved image by measuring the accuracy compared with the value of accuracy with previous methods.
The main objective of this paper project was to create a state-of-the-art face identification technique that can handle the various difficulties caused by changes in illumination, occlusions, and facial emotions. Face detection is a cornerstone of computer vision, facilitating diverse applications ranging from surveillance systems to human-computer interaction. Throughout this paper, the comprehensive exploration of advancing face detection methodologies has been undertaken, culminating in developing and evaluating a novel approach. The challenges posed by variations in facial expressions, lighting conditions, and occlusions necessitated a multifaceted solution. Our proposed method, which consists of interconnected steps, works quite well to overcome these challenges. Using deep learning architectures to increase feature extraction and discrimination was beneficial in the initial stage of fine-tuning Residual Networks (ResNet-50) to serve as the Region-based Convolutional Neural Network (Faster R-CNN) framework classifier. The process of gradually optimizing thresholds, such as batch size, learning rate, and detection threshold, involved using the Gray Wolf optimization technique (GWO). The conversion process was accelerated and improved overall detection process efficiency and accuracy using a clever fusion of machine learning and metaheuristic optimization techniques. A key component of our methodology is the careful data processing, which was necessary to ensure. The suggested method was carefully examined on a particular dataset, and the 94% training accuracy that was attained together with an identical test dataset accuracy highlights the method’s resilience. These findings support the effectiveness of our approach in reducing false positives and negatives, resulting in unmatched recall and precision in the detection system. The discovery has significant significance as it can potentially improve face detection systems’ performance and reliability in various real-world applications, such as human-computer interaction and surveillance. Convolutional neural networks, deep learning architectures, and metaheuristic optimization approaches were synergized to produce a new and reliable solution
The main objective of this paper project was to create a state-of-the-art face identification technique that can handle the various difficulties caused by changes in illumination, occlusions, and facial emotions. Face detection is a cornerstone of computer vision, facilitating diverse applications ranging from surveillance systems to human-computer interaction. Throughout this paper, the comprehensive exploration of advancing face detection methodologies has been undertaken, culminating in developing and evaluating a novel approach. The challenges posed by variations in facial expressions, lighting conditions, and occlusions necessitated a multifaceted solution. Our proposed method, which consists of interconnected steps, works quite well to overcome these challenges. Using deep learning architectures to increase feature extraction and discrimination was beneficial in the initial stage of fine-tuning Residual Networks (ResNet-50) to serve as the Region-based Convolutional Neural Network (Faster R-CNN) framework classifier. The process of gradually optimizing thresholds, such as batch size, learning rate, and detection threshold, involved using the Gray Wolf optimization technique (GWO). The conversion process was accelerated and improved overall detection process efficiency and accuracy using a clever fusion of machine learning and metaheuristic optimization techniques. A key component of our methodology is the careful data processing, which was necessary to ensure. The suggested method was carefully examined on a particular dataset, and the 94% training accuracy that was attained together with an identical test dataset accuracy highlights the method’s resilience. These findings support the effectiveness of our approach in reducing false positives and negatives, resulting in unmatched recall and precision in the detection system. The discovery has significant significance as it can potentially improve face detection systems’ performance and reliability in various real-world applications, such as human-computer interaction and surveillance. Convolutional neural networks, deep learning architectures, and metaheuristic optimization approaches were synergized to produce a new and reliable solution.
Brain tumors are collections of abnormal tissues within the brain. The regular function of the brain may be affected as it grows within the region of the skull. Brain tumors are critical for improving treatment options and patient survival rates to prevent and treat them. The diagnosis of cancer utilizing manual approaches for numerous magnetic resonance imaging (MRI) images is the most complex and time-consuming task. Brain tumor segmentation must be carried out automatically. A proposed strategy for brain tumor segmentation is developed in this paper. For this purpose, images are segmented based on region-based and edge-based. Brain tumor segmentation 2020 (BraTS2020) dataset is utilized in this study. A comparative analysis of the segmentation of images using the edge-based and region-based approach with U-Net with ResNet50 encoder, architecture is performed. The edge-based segmentation model performed better in all performance metrics compared to the region-based segmentation model and the edge-based model achieved the dice loss score of 0. 008768, IoU score of 0. 7542, f1 score of 0. 9870, the accuracy of 0. 9935, the precision of 0. 9852, recall of 0. 9888, and specificity of 0. 9951.
Deep falsification of multimedia content, especially videos and photos, threatens social cohesion (e.g., rumour propagation, extortion, and truth distortion) and must not be ignored. In some cases, this issue requires effective detection solutions. Most studies suggest that convolutional neural networks (CNNs) may not be able to extract complex features like those used in deepfake production. Thus, hybrid approaches that can capture complex features and act as powerful descriptors for binary classification are needed to separate bogus from true content. In this paper, a hybrid algorithm is developed to combine gated recurrent units (GRU) and CNN. The proposed model aims to improve the extraction of complex features by simultaneously capturing instantaneous and spatial features. This approach permits the extraction of implicit features that are vital to the final classification process, especially when dealing with a sequential series within video content. Finally, a dense neural network is used to classify these features. Practically, two data sets were used to train the proposed model: the FaceForensics++ (FF++) and DeepFake Detection Challenge (DFDC) datasets. The evaluation results of the proposed model on the FF++ dataset for the Area Under the Curve (AUC) and F1-score metrics reached 0.88% and 0.85%, respectively. While DFDC is 0.95% and 0.86% for the same metrics, respectively.
Face recognition is the technology that verifies or recognizes faces from images, videos, or real-time streams. It can be used in security or employee attendance systems. Face recognition systems may encounter some attacks that reduce their ability to recognize faces properly. So, many noisy images mixed with original ones lead to confusion in the results. Various attacks that exploit this weakness affect the face recognition systems such as Fast Gradient Sign Method (FGSM), Deep Fool, and Projected Gradient Descent (PGD). This paper proposes a method to protect the face recognition system against these attacks by distorting images through different attacks, then training the recognition deep network model, specifically Convolutional Neural Network (CNN), using the original and distorted images. Diverse experiments have been conducted using combinations of original and distorted images to test the effectiveness of the system. The system showed an accuracy of 93% using FGSM attack, 97% using deep fool, and 95% using PGD.
Automatic signature verification methods play a significant role in providing a secure and authenticated handwritten signature in many applications, to prevent forgery problems, specifically institutions of finance, and transections of legal papers, etc. There are two types of handwritten signature verification methods: online verification (dynamic) and offline verification (static) methods. Besides, signature verification approaches can be categorized into two styles: writer dependent (WD), and writer independent (WI) styles. Offline signature verification methods demands a high representation features for the signature image. However, lots of studies have been proposed for WI offline signature verification. Yet, there is necessity to improve the overall accuracy measurements. Therefore, a proved solution in this paper is depended on deep learning via convolutional neural network (CNN) for signature verification and optimize the overall accuracy measurements. The introduced model is trained on English signature dataset. For model evaluation, the deployed model is utilized to make predictions on new data of Arabic signature dataset to classify whether the signature is real or forged. The overall obtained accuracy is 95.36% based on validation dataset.
Many assistive devices have been developed for visually impaired (VI) person in recent years which solve the problems that face VI person in his/her daily moving. Most of researches try to solve the obstacle avoidance or navigation problem, and others focus on assisting VI person to recognize the objects in his/her surrounding environment. However, a few of them integrate both navigation and recognition capabilities in their system. According to above needs, an assistive device is presented in this paper that achieves both capabilities to aid the VI person to (1) navigate safely from his/her current location (pose) to a desired destination in unknown environment, and (2) recognize his/her surrounding objects. The proposed system consists of the low cost sensors Neato XV-11 LiDAR, ultrasonic sensor, Raspberry pi camera (CameraPi), which are hold on a white cane. Hector SLAM based on 2D LiDAR is used to construct a 2D-map of unfamiliar environment. While A* path planning algorithm generates an optimal path on the given 2D hector map. Moreover, the temporary obstacles in front of VI person are detected by an ultrasonic sensor. The recognition system based on Convolution Neural Networks (CNN) technique is implemented in this work to predict object class besides enhance the navigation system. The interaction between the VI person and an assistive system is done by audio module (speech recognition and speech synthesis). The proposed system performance has been evaluated on various real-time experiments conducted in indoor scenarios, showing the efficiency of the proposed system.
Video prediction theories have quickly progressed especially after a great revolution of deep learning methods. The prediction architectures based on pixel generation produced a blurry forecast, but it is preferred in many applications because this model is applied on frames only and does not need other support information like segmentation or flow mapping information making getting a suitable dataset very difficult. In this approach, we presented a novel end-to-end video forecasting framework to predict the dynamic relationship between pixels in time and space. The 3D CNN encoder is used for estimating the dynamic motion, while the decoder part is used to reconstruct the next frame based on adding 3DCNN CONVLSTM2D in skip connection. This novel representation of skip connection plays an important role in reducing the blur predicted and preserved the spatial and dynamic information. This leads to an increase in the accuracy of the whole model. The KITTI and Cityscapes are used in training and Caltech is applied in inference. The proposed framework has achieved a better quality in PSNR=33.14, MES=0.00101, SSIM=0.924, and a small number of parameters (2.3 M).
Global agriculture employs central pivot irrigation system(CPIS) as a highly significant method for intelligent irrigation. Cultivating crucial crops like wheat and other strategically important crops that occupy extensive land areas contributes to global food security. The Central Pivot Irrigation System encounters technical issues that result in malfunctions in its automatic control system. These malfunctions occasionally cause damage to the primary pipes and towers that operate the system, resulting in significant material losses for farmers and agricultural crops. Moreover, the repair process is time-consuming. Therefore, to address this issue, this study employed the YOLOv5 models to accurately identify and detect defects in the CPIS machine by determining whether they are in a safe or dangerous state. The dataset that was used in this study was gathered from agricultural areas in Salah al-Din Governorate. The CPIS detection model yielded the following results: the grayscale color system with Yolov5n achieved a 98 % detection rate with accuracy and F1-score values of 0.866. Similarly, Yolov5m achieved a 98 % detection rate with accuracy and F1-score values of 0.804. In the RGB color system, the maximum results achieved with Yolov5n are 97 % for accuracy and 0.812 for F1-score. On the other hand, Yolov5s6 achieves a result of 95 % for accuracy and 0.82 for both F1-score and accuracy. Based on the aforementioned outcome, we can infer that yolov5s6 accurately detects the CPIS in both its safe and dangerous states. Therefore, they can be deployed in a real-time system for CPIS defect monitoring and control systems.
Alzheimer’s disease (AD), the most common form of dementia, affects over 55 million people worldwide. The most form of dementia progresses into three distinct stages: mild, moderate, and very mild compared to Cognitively Normal (CN). Early detection is crucial to prevent brain damage before the late stages. Convolutional Neural Networks (CNNs), a subfield of deep learning, have recently found remarkable applications in medical image processing and computer-aided diagnosis (CAD). To this end, this paper presents a new efficient multi-classification AlzCNN-Net model to enhance the accuracy and efficacy of MRI image classification for various Alzheimer’s disease conditions. Initially, the training process involves utilizing open-source Alzheimer’s disease datasets from the Kaggle database to classify the brain MRI into its corresponding category. To verify the model’s efficacy, a comparative analysis with three pre-trained models, namely VGG16, Incep-tionV3, and MobileNetV2, has been investigated via transfer learning applied to the same dataset. As a result, the findings reveal that the AlzCNN-Net model exhibits an optimal performance, attaining the best accuracy in training with 99.67%, validation with 98.24%, and testing with 98.9% accuracy at epoch 100 with batch size 32 compared to the existing pre-trained approaches.
In smart cities, health care, industrial production, and many other fields, the Internet of Things (IoT) have had significant success. Protected agriculture has numerous IoT applications, a highly effective style of modern agriculture development that uses artificial ways to manipulate climatic parameters such as temperature to create ideal circumstances for the growth of animals and plants. Convolutional Neural Networks (CNNs) is a deep learning approach that has made significant progress in image processing. From 2016 to the present, various applications for the automatic diagnosis of agricultural diseases, identifying plant pests, predicting the number of crops, etc., have been developed. This paper involves a presentation of the Internet of Things system in agriculture and its deep learning applications. It summarizes the most essential sensors used and methods of communication between them, in addition to the most important deep learning algorithms devoted to intelligent agriculture.