This paper presents a bidirectional AC-DC converter system designed for seamless power exchange between electric vehicles (EVs) and the utility grid. The proposed converter facilitates the conversion of 230 V, 50 Hz AC input to 380 V DC during grid-to-vehicle operation, allowing for efficient battery charging through a bidirectional DC-DC converter. Conversely, during vehicle-to-grid operation, it converts the 380 V DC input from the DC-DC converter to 230 V, 50 Hz AC output for grid supply. The system employs PI controllers to ensure precise voltage and current regulation, ensuring stable and efficient operation during grid interaction. Simulation results demonstrate the system’s effectiveness in managing power conversion for both grid-to-vehicle (G2V) and vehicle-to-grid (V2G) applications This paper presents a bidirectional AC-DC converter system designed for seamless power exchange between electric vehicles (EVs) and the utility grid. The proposed converter facilitates the conversion of 230 V, 50 Hz AC input to 380 V DC during grid-to-vehicle operation, allowing for efficient battery charging through a bidirectional DC-DC converter. Conversely, during vehicle-to-grid operation, it converts the 380 V DC input from the DC-DC converter to 230 V, 50 Hz AC output for grid supply. The system employs PI controllers to ensure precise voltage and current regulation, ensuring stable and efficient operation during grid interaction. Simulation results demonstrate the system’s effectiveness in managing power conversion for both grid-to-vehicle (G2V) and vehicle-to-grid (V2G) applications

Facial recognition systems are widely used in various applications such as security, healthcare, and authentication, but face significant challenges in uncontrolled environments. Poor lighting conditions can obscure facial features, introduce shadows, and distort spatial relationships, while changes in pose are critical for accurate identification. Existing methods often struggle to strike a balance between accuracy, computational efficiency, and robustness. Deep learning has become popular for automatically learning features through convolution layers. This study proposes a robust framework that integrates contrast-limited adaptive histogram equalization (CLAHE) and adaptive gamma correction for illumination normalization and multi-task cascaded convolutional networks (MTCNN) for precise face detection under varying poses and lighting conditions. This study proposes a deep learning-based approach for face recognition utilising multiple models, including VGG16, VGG19, ResNet-50, ResNet-101, and MobileNetV2. For classification, an ensemble of SVM, XGBoost, and random forest classifiers is combined using weighted averaging. The approach is tested on datasets such as CASIA3D and 105PinsFace, which include variations in illumination conditions. Using deep learning for automated hierarchical feature extraction and ensemble strategies, experimental results demonstrate significant improvements in recognition accuracy and enhanced robustness against lighting and pose variations while ensuring scalability for real-world applications. The approach achieved 99.91% accuracy on the CASIA3D dataset and 98.77% on the 105PinsFace dataset, showcasing its effectiveness across challenging conditions.

This paper presents a comparative analysis of YOLOv11 and YOLOv12 for automated weed detection in precision agriculture. The primary objective is to assess both models' detection accuracy, generalization ability, and reliability using a custom-annotated dataset of sesame crop and weed images. YOLOv11, known for its faster inference speed, demonstrates higher mAP@0.5 in straightforward detection scenarios. However, YOLOv12 outperforms in challenging conditions due to its advanced architectural enhancements, including attention mechanisms and improved feature pyramids. This study highlights the trade-off between computational efficiency and robust detection, offering insights into choosing the optimal object detection model for real-time agricultural applications

In distributed computing environments, computation offloading is a vital strategy for maximizing the performance and energy efficiency of mobile devices. Distributed deep learning-based offloading (DDLO) [10] and deep reinforcement learning for online computation offloading (DROO) [10] are two popular methods for solving the computation offloading problem. In DDLO, the data is divided into smaller pieces during offloading and distributed throughout the systems or devices. In DROO, an agent is trained to determine the optimum offloading choices based on the resources at hand, the network environment, and the application's performance requirements. Comparison is presented of both approaches, emphasizing their benefits and drawbacks and the situations when one approach is more suitable than the other. Precision, effectiveness, and adaptability are just a few of the different metrics we use to evaluate the performance of both techniques in a variety of workload and network configuration scenarios. Our findings indicate that while deep reinforcement learning is more able to respond to environmental changes, distributed deep learning-based offloading is more efficient in terms of computational resources.

Facial image analysis and categorization have recently made great strides in computer vision. The current study, explores ways to help computers better recognize faces quickly and accurately, especially for tasks like security and entertainment. Identifying faces, emotions, and identities is crucial in Security and Surveillance, Access Control, user Authentication in Smart Devices, and Emotion Analysis in Human-Computer Interaction. Adopting the MobileNet deep learning model because it requires less memory works efficiently. To make it even more effective at recognizing faces, adjusted its parameters and tested it with two data sets, the CASIA 3D face data set and 105 pins data set. The study using MobileNetV2 achieved a very high accuracy of 98.71% on the CASIA 3D face data set and 99.29% on the 105 pins data set. The experimental results show that MobileNetV2 better understands faces in different situations.

The demand for food production has led to advancements in precision agriculture, aiming to enhance crop yield and quality. This study investigates the application of deep learning algorithms, including GoogLeNet, RESNET-50, MobileNet-v2, VGG-16, and ShuffleNet, for automated plant disease detection. The research utilizes a dataset comprising images of citrus diseases to train and evaluate the models. Results show promising accuracy rates, highlighting the potential of deep learning in optimizing resource utilization and facilitating timely interventions in agriculture.

Lung cancer represents a significant contributor to global cancer-related deaths, underscoring the critical need for early detection to reduce mortality rates. Using convolutional neural networks (CNNs) and deep learning, with a specific focus on MobileNet, VGG16, GoogleNet, InceptionentV3 and ResNet50, this study delves into the integration of AI for lung cancer detection using the LC25000 dataset, encompassing a diverse range of lung pathology CT scans. By tailoring the MobileNet architecture and optimising it for CT image analysis, the research strives to enhance the model's precision in identifying lung malignancies. The customized MobileNet, InceptionNet, GoogleNet, ResNet50, VGG16 model undergoes fine-tuning via strategic adjustments and training to discern subtle patterns indicative of lung cancer. Ensembled with these models to give accurate results. within medical imaging datasets. Robust evaluation techniques are implemented to gauge the model's efficacy, incorporating metrics such as accuracy and computational efficiency, positioning it as a promising tool for advancing early lung cancer detection methodologies. © 2024 IEEE.

Biometric template protection is important in the current situation because of the reliance on intelligent approaches for recognizing an individual. In recent years, there have been numerous high-profile frauds. Texture, color, and shape are the most prevalent biometric features employed today. Insufficient user data and compromised keys cast doubt on the dependability of biometric systems. The proposed system employs a novel technique to close this gap. It extracts biometric features from a person’s face, iris, and palmprint. Combining biometric features increases system reliability, safety, and user privacy. We used a colorization technique to generate three separate color Quick Response (QR) codes from a user-defined Red Green Blue (RGB) tuple random seed, which we then combined to create a one-way cancellable template. This work provides a biometric verification system that supports numerous cancellation mechanisms. This would reduce the dangers associated with biometric templates, confusing systems, user concerns, and fraud. The database saves these templates for future user-driven security key-based situations involving intra-class and inter-class verification in the changed domain. The system’s remarkable performance resulted in improved accuracy and security, reaching 99.84% overall with a 0.11% crossover error rate at an optimal threshold of 5.59. Finally, the Area Under the Receiver Operating Curve (AU_ROC) was 0.97, which is closer to the optimal value of 1.

The recommendation system is one of the most essential information services in today's online business applications, such as Amazon, Flipkart, and YouTube. In recent days, deep learning and machine learning models have performed exceedingly well in various applications related to text processing, image processing, audio and video processing. This work aims to review recent studies that evaluated behavior of various deep learning and machine learning models in recommender systems, and summarize the various key insights related to their performance in these applications. Specifically, we focus on analysis of four types of deep learning and machine learning techniques: graph-based baselines, sequential baselines, self-supervised sequential models, and self-supervised graph-based models. Moreover, these models are evaluated on four different types of datasets: Yelp 2018, Ml-1M, Amazon Beauty, and iFashion. Among the eleven different models employed for this analysis, the two self-supervised sequential models, CL4SRec and BERT4Rec, outperform in terms of two of the four distinct metrics (Recall and NDCG) used.

Recent years have seen increased interest in research on biometric template protection due to the widespread use of biometric authentication systems. For providing security to biometric templates, there is a procedure known as Fully Homomorphic Encryption that protects biometric templates from the malicious server environment. Users have more customization options with a biometric authentication system than a password or token. The most popular biometric modalities include fingerprints, iris scans, facial images, etc. Biometric modalities offer security on IoMT -based systems. Face recognition is one of the most often utilized biometric authentication methods in societal structure. Face recognition technology has made enormous strides in recent years. Here, we examine the viability of securing a database of iris templates using a methodology based on fully homomorphic Encryption. By directly matching templates in the encrypted domain, this framework is designed to protect confidentiality and restrict information from leaking from the templates while preserving their utility. We also investigate various classification techniques on machine learning models to achieve improved accuracy with shorter execution times. The aggregate verification vector assists in confirming the accuracy of the computed classification result, and the CKKS technique ensures confidentiality for the biometric templates. This study provides a plethora of information on fully homomorphic biometric authentication, containing a wide assortment of algorithms that satisfy homomorphic Encryption and various methods for extracting the biometric-related template.

With the enormous increase in the number of vehicles that are making use of roadways day by day, traffic congestion is one of the significant issues that are being observed. This problem can be addressed with proper traffic regulation. In this paper, proposing an automated system that will perform traffic analysis from the traffic videos which were captured from static cameras. This traffic analysis will be performed in three stages: vehicle detection, vehicle classification, into ten major categories, followed by vehicle counting under each category. The proposed work adopting the background subtraction method and vehicle classification by using a pre-trained model VGG16 as well as logistic regression (LR). Observations were made across several models ranging from the models that were built from scratch to pre-trained models as well as VGG16 + logistic regression. Experimental results show that the proposed model provides top-1 accuracy of 84.32% and top-5 accuracy of 99.77%.

There are many types of fraud in our daily life. One of the frauds occurring these days is credit card fraud. When people around the globe make credit card transactions, there will also be fraudulent transactions. To avoid credit card fraud, we must know the patterns and how the fraud values differ. This paper proposed credit card fraud detection using machine learning based on the labeled data and differentiating the fraudulent and legitimate transactions. The experiment was conducted using supervised machine-learning techniques.

Face recognition technology is one of the everyday tasks in our daily life. But, recognising the correct face with high accuracy from large databases is a challenging task. To overcome this challenge, feature fusion of local binary pattern (LBP) with Gabor-Kernel Fisher analysis (Gabor-KFA) has proposed for face recognition. In this method, by using Gabor filter, extract Gabor features from a face image, on the other hand, extract features from LBP coded face image, then combined these extracted features generate high dimensional feature space. With this high dimensionality features, the complexity of training time and identification time may increase. To avoid this complexity, the Kernel Fisher analysis algorithm was adopted to reduce the feature vector size. Experiments were conducted separately on Gabor features and also on fused features. To test the performance of the proposed approach, the experiments were performed on the IIT Delhi database, ORL database, and FR database.

Social webs like Instagram, Twitter, and WhatsApp are full of deliberations involving sentiments, feelings, and impressions of human beings worldwide. Moreover, understanding and segregating texts based on emotions is a complex task that could be considered progressive sentiment analysis. As sentiments play a crucial role in human interaction, the skills to perceive it through textual content analysis has numerous applications in natural language processing (NLP) and human–computer interaction (HCI). This paper suggests classifying and examining tweets based on six basic emotions: happiness, fear, anger, disgust, surprise, and sadness. Language translators are used to apply data augmentation. Experimental results show that augmented data provides better results than the original data.

-

A public health emergency threat is happening due to novel coronavirus 2019 (nCoV-2019) throughout the world. nCoV-2019 is also named Severe Acute Respiratory SyndromeCoronaVirus-2 (SARS-CoV-2). COVID-19 is the disease caused by this virus. The virus originates in bats and is transmitted to humans by some unidentified intermediate animals. This virus started around December 2019 at Wuhan of China. After that, it turned into a pandemic. Even though there is no efficient vaccination, the entire world fights against the COVID-19. This article presents an overview of the scenario of the world as well as India. Some of the leading countries in the world are also affected by this virus badly. Even India is the 2nd highest population, is taking necessary precautions to protect it. With the Government of India's decisions, along with effective social distancing and hygienic measures, India is in a better position. But, in the future, COVID19 cases in India, still unpredictable. We designed an algorithm based on Convolutional Neural Network (CNN), which helps to classify COVID19+ and COVID19- persons using people's chest X-ray images automatically generated within the shortest time. The proposed method discovered that employing CT scan medical images produced more accurate results than X-ray images.