Browse Items (3095 total)
Sort by:
-
Exploring the Nexus of Deepfakes and VFX Technology: Unveiling Insights, Challenges, and Innovations
This research paper explores the intersection of Deepfakes and Visual Effects (VFX) technology, investigating their convergence, implications, and advancements. Deepfakes, driven by artificial intelligence algorithms, have revolutionized the creation of synthetic media, while VFX techniques have long been utilized in the film industry for various purposes. This paper delves into the technical underpinnings of both Deepfakes and traditional VFX, highlighting similarities, differences, and synergies. It examines the potential applications of Deep-fakes in VFX-driven storytelling, digital compositing, and character animation, while also addressing the ethical concerns and risks associated with their misuse. Furthermore, the paper discusses emerging trends and innovations that bridge the gap between Deepfakes and VFX technology, paving the way for new creative possibilities and challenges. Through a comprehensive analysis, this paper aims to provide valuable insights into the evolving landscape of synthetic media and its implications for the VFX industry. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Concernment of Feature Selection Using Classification Algorithms and Developing the Web Frame for Breast Cancer Prediction
Breast cancer is invasive cancer and it is the most common cancer diagnosed in women. The survival rate of breast cancer patients is increasing due to timely detection, better empathy about the disease, and new tailored approach for the treatment. Even hormonal imbalance, environmental factors, gene mutation, and lifestyle are also the reasons for breast cancer. Stages of breast cancer majorly depend on the size of the tumor as well as the spreading of cancer to the lymph nodes. An instinctive disease detection system and computer-aided diagnosis will help the medical practitioners in early prediction of breast cancer using machine learning algorithms. In this paper, Random Forest for ranking the features by assigning the weights and selection of features using support vector machine and Nae Bayes are used. The Breast Cancer Wisconsin Dataset from the UCI Repository has been taken for examination purposes. Features selected from support vector machine and Naive Bayes have been tested by using seven different classifiers: logistic regression, random forest, K-nearest neighbor, support vector classifier, linear support vector classifier, Gaussian Naive Bayes, and decision tree. Based on the experimental results with 7030 and 8020 splits, 7030 is obtained with the best accuracy. Support vector machine with 12 features resulted in an accuracy of 97.66% and Nae Bayes with 17 features resulted in an accuracy of 96.49% with the improved results as compared to without feature selection. As support vector machine resulted with best accuracy with 12 features, by using these 12 features, web application for the prediction of breast cancer has been developed using Web framework using Python Flask, PyCharm IDE, and the instance has been executed virtually in the Amazon EC2 cloud Platform. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Development of Enhance-Net Deep Learning Approach for Performance Boosting on Medical Images
Only a few clinical procedures include the use of clinical methods for the early detection, observing, evaluation, and treatment evaluation of a range of medical illnesses. Knowing the analysis of medical images in computer vision necessitates being acquainted with the core concepts and uses of deep learning and artificial neural networks. The A rapidly expanding area of study is the Deep Learning Approach (DLA) in medical image processing. DLA is often used in medical imaging to determine if an ailment is present or not. By producing speedier, more accurate results in real time, deep learning algorithms may make the jobs of radiologists and orthopaedic surgeons easier. But the standard deep learning approach has reached its efficiencies. While offering an ideal solution known as boost-Net, we study numerous optimization strategies to increase the effectiveness of deep neural networks in this research. From a selection of well-known deep learning models, Champion-Net was selected as the deep learning model. The musculoskeletal radiograph-bone classification (MURA-BC) dataset is used in this investigation. Utilizing the train and test datasets, Enhance-Net's classification precision was evaluated. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Digitization of Monuments An Impact on the Tourist Experience with Special Reference to Hampi
The cultural heritage of India offers a deep examination of the country's political and historical evolution. Historical structures and monuments are among a nation's most valuable assets and a source of pride for Indian civilization. Monuments hold significant historical importance and exert a profound emotional influence on the community. Given the deterioration of culturally significant heritage monuments caused by factors such as weather, climate change, and human activity, as well as the threats these elements pose to numerous heritage sites of national and international significance, it is imperative to prioritize the recording, preservation, and conservation of these monuments. Events of cultural significance require comprehensive digital documentation and proper recording. As demonstrated by various programs and initiatives led by Prime Minister Narendra Modi, the government is committed to enhancing the visitor experience at monuments and museums. The primary aim of the current study is to better understand how cultural heritage sites are digitized and to assess the implications of this process for enhancing the tourist experience. To address the research objectives, a survey was conducted to analyze digital requirements. The digitization of significant cultural heritage sites is vital for the long-term sustainability of the tourism industry. Many methods will be adapted as resources permit, ensuring the industry's steady growth over time. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Implementation of Recent Advancements in Cyber Security Practices and Laws in India
In the past few decades, a large number of scholars and experts have found that wireless connectivity technologies and systems are susceptible to many kinds of cyber attacks. Both governmental organizations and private firms are harmed by these attacks. Cybersecurity law is a complex and fascinating area of law in the age of information technology. This essay aims to outline numerous cyber hazards as well as ways to safeguard against them. In both local and international economic contexts, it is critical to establish robust regulatory and legal structures that address the growing concerns about fraud on the internet, security of information, and intellectual property protection. Additionally, it covers cybercrime's different manifestations and security in a global perspective. Due to recent technical breakthroughs and a growth in access to the internet, cyber security is now utilized to safeguard not just a person's workstation but also their own mobile devices, including tablets and mobile phones, that have grown into crucial tools for data transmission. The community of security researchers, which includes members from government, academia, and industry, must collaborate in order to comprehend the new risks facing the computer industry. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
2D Photonic Crystal for the Detection of Infectious Virus and Bacterial Diseases
In this paper, photonic crystal (PhC) sensor for analysis and modelling for viral and bacterial detection is proposed. Optical biosensors detect cancer, Bacillus cereus, malaria, typhoid, tuberculosis, etc. Optical biosensors work by shifting the peak resonance wavelength with modest refractive index changes. Because viral pathogens rapidly mutate and replicate in the human cell nucleus, sensors that offer accurate results for viral and bacterial diseases in seconds are in high demand. Hence, optical biosensors provide fast, sensitive results. The sensor detects influenza H1N1, hepatitis B (HBV), and typhoid, respectively. A maximum sensitivity of 443.33nm/RIU with a quality factor of 1309 is obtained. Simulations are performed using finite-difference time-domain (FDTD). The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
An Enhanced Whale Optimization Algorithm for Task Scheduling in Cloud Computing
Task Scheduling is the significant challenge in the environment of Cloud Computing (CC) and has attention in numerous researchers in recent years with respect to attain cost effective computation and improve resource utilization. The existing algorithms has limitations of role and selection criteria of inertia weight was not considered. In this research, Enhanced Whale Optimization Algorithm (EWOA) is proposed for maximize effectiveness of task scheduling in CC. An inertia weight is implemented in WOA algorithm that enhances the convergence and accuracy of algorithm that helps in task scheduling effectiveness. The performance of proposed technique is estimated with performance measure of Makespan (ms), execution time (s) and resource utilization (%). The proposed method attained less execution time of 2304, 2537, 2765, 2983 and 3016s for 200, 400, 600, 800 and 1000 number of tasks. The proposed method attained the superior results when compared with other existing algorithms like Ant Colony Optimization (ACO), Grey Wolf Optimization (GWO), Particle Swarm Optimization (PSO) and Whale Optimization Algorithm (WOA). The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Multi-objective Deep Reinforcement Learning Approach for Multiple -Input/Multiple-output Routing in WSN
The Wireless Sensor Network (WSN) is a network of numerous devices that are interconnected via the internet and significantly impact the network. However, despite their significant applications WSNs face challenges related to network security energy levels and information transmission delays. To address these challenges, a method utilizing Multi-Objective Deep Reinforcement Learning (DRL) has been proposed. The proposed method aims to maximize energy utilization in the network by efficiently managing covered and uncovered cluster network routing. The performance of energy transmission is enhanced through the use of the Markov Decision Process model based on multi-objective DRL combined with training the network using Deep Q Network (DQN) to reduce network energy consumption. Training the network with multiple objectives may pose challenges requiring more samples and leading to higher sample complexity, which can be a limiting factor in real-world applications. Despite this, the proposed multi-objective DRL method demonstrates high performance compared to existing methods such as Particle Swarm Optimization (PSO) and Convolutional Neural Network (CNN). Specifically, multi-Objective DRL method yields superior results, achieving an energy consumption of 42J, Packet Delivery Ratio (PDR) of 90%, and an End-To-End Delay (ETED) of 45 S. These outcomes outperform existing methods in the context of WSNs. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
RehabPal: Automated Physical Therapy Assistance
This innovative research project transforms automated physical therapy support by combining 2D and 3D pose assessment approaches with tailored feedback systems. Utilizing Mediapipe technology, the apparatus attentively observes patients movements in a two-dimensional space and gives realtime data regarding their gait, exercise compliance, and range of motion. Accurate data on joint angles and body segment alignment are provided by the system's sophisticated 3D posture estimation algorithms, which enhance tracking precision. The system incorporates customized feedback systems that include individual patients goals, progress, and conditions and delivery and reward schemes. By increasing user engagement and adherence, the integration of gamification elements has the potential to revolutionize automated physical therapy support. This comprehensive approach aims to enhance patients quality of life while simultaneously enhancing long-term rehabilitation outcomes by providing more conveniently accessible, affordably priced, and specially designed therapies. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Deploying Deep Learning in Real-Time for Lung Cancer Diagnosis via Medical Imaging
In this research, deep learning models were used to diagnose lung cancer automatically using hospital image data. A dataset with 3,400 lung cancer images from online repositories and hospital archives was used for model training and evaluation. After preprocessing and feature extraction, various deep learning architectures such as VGG-16, CNN, ResNet and RNN were adopted in this study. The VGG-16 model had the highest accuracy rate of 96.86%, showing strong performance. This rate of accuracy is actually higher than their accuracy of 91%. These results serve to highlight the impressive accuracy achieved by our study relative to prior research. By accurately and effectively altering lung cancer diagnosis into a process entirely reliant on algorithms, deep learning models show promise for their potential. Diagnostic tools should be able to catch cancer early and accurately, identify the present type and classification for tumors. For all its promise, limitations such as dataset size and generalizability mean that clinical trials will be needed for further validation. Focus should turn toward this as the direction of future research in order to enhance model robustness and applicability against challenges. This research allows us to better the well-being of patients and reduce the burden of lung cancer through timely intervention and personalized treatment strategies by making use of advanced techniques in medical diagnostics. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Enhancing Diagnostic Accuracy for Autism with BRCNet: A Novel Approach for Brain Region Segmentation and Classification Using Deep Learning
In the quest to enhance the diagnostic accuracy of neural disorders, particularly autism, this paper presents a novel approach for brain region classification using advanced machine learning techniques. The study utilizes the ABIDE and AAL116 atlas datasets, focusing on segmenting and classifying brain regions from resting-state functional MRI (rs-fMRI) images. We propose a three-stage process. In the first stage, data collection and preprocessing are conducted, where rs-fMRI images are preprocessed into SPM12-NIfTI format. The second stage involves the segmentation of brain regions using a Regularized VNet, resulting in the extraction of AAL116 brain region images, which are then split into training, testing, and validation sets. In the third stage, we introduce a custom-designed BRCNet (Brain Region Classification Network), which discriminates between Autism and Normal classes. Our segmentation methods are rigorously evaluated using metrics such as Dice Score, Recall, and Precision, with the proposed method achieving a Dice Score of 0.985, Recall of 0.962, and Precision of 0.991, surpassing other tested methods like UNet, Active Contour, and Binary Unit. For classification, various methods, including Support Vector Machines (SVM), Decision Trees (DT), and Neural Networks like ResNet, are compared. Our findings demonstrate that ResNet achieves an exemplary performance with an Accuracy of 97.5%, Sensitivity of 96.2%, Specificity of 97.1%, Precision of 97%, and an F-Measure of 97.93%. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Prioritisation of Challenges in OTT Video Platforms: A Multi-criteria Decision-Making Approach
Over-the-top (OTT) video platforms have emerged as the preferred choice for on-demand entertainment. In this ever-changing landscape, OTT video platforms face many challenges to being relevant in the market. Identifying and prioritizing challenges is pivotal for the sustainable growth of OTT platforms. This paper aims to comprehensively examine and prioritize the challenges of the OTT video platform. The challenges are identified through an extensive literature review and unstructured interviews with six OTT industrial experts. The importance of each challenge is measured based on the analytical hierarchy process (AHP) to develop a hierarchy of those challenges. The AHP analysis results indicated customer retention as the most significant challenge, followed by content, customer experience, infrastructure, and bandwidth. The study is subjective to the experts opinions and available literature regarding the OTT platforms. The insights gleaned from this research are poised to offer substantial value to digital platform operators, media professionals, and managers. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Prioritisation of Challenges in OTT Video Platforms: A Multi-criteria Decision-Making Approach
Over-the-top (OTT) video platforms have emerged as the preferred choice for on-demand entertainment. In this ever-changing landscape, OTT video platforms face many challenges to being relevant in the market. Identifying and prioritizing challenges is pivotal for the sustainable growth of OTT platforms. This paper aims to comprehensively examine and prioritize the challenges of the OTT video platform. The challenges are identified through an extensive literature review and unstructured interviews with six OTT industrial experts. The importance of each challenge is measured based on the analytical hierarchy process (AHP) to develop a hierarchy of those challenges. The AHP analysis results indicated customer retention as the most significant challenge, followed by content, customer experience, infrastructure, and bandwidth. The study is subjective to the experts opinions and available literature regarding the OTT platforms. The insights gleaned from this research are poised to offer substantial value to digital platform operators, media professionals, and managers. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Predicting Financial Distress in India: A Deep Learning Approach
The present study examines the efficacy of deep learning models in predicting financial distress in India. For this purpose, the study employs three distinct architectures: Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), and Conventional Neural Network (CNN) models. Utilizing data from companies that filed for bankruptcy under the Insolvency and Bankruptcy Code 2016 for the period of 20162023, the study adopts a balanced sample approach to categorize them into distressed and non-distressed groups. Nineteen financial variables are utilized to predict financial distress. Python is used as the programming language, and Jupyter Notebook facilitates algorithm development. The findings reveal that the LSTM model, when compared to RNN and CNN, achieved 91% accuracy using parameters such as 8 LSTM units with tanh activation and a dense layer with sigmoid activation function, a batch size of 10, 50 epochs, RMSprop optimizer, and binary cross-entropy loss were used. The study suggests that deep learning presents a novel approach that can enhance performance in financial distress prediction studies. This study is believed to be the first to utilize deep learning models for financial distress prediction in India based on single-year data, offering valuable insights for financial institutions and investors seeking more effective risk management strategies. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Predicting Financial Distress in India: A Deep Learning Approach
The present study examines the efficacy of deep learning models in predicting financial distress in India. For this purpose, the study employs three distinct architectures: Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), and Conventional Neural Network (CNN) models. Utilizing data from companies that filed for bankruptcy under the Insolvency and Bankruptcy Code 2016 for the period of 20162023, the study adopts a balanced sample approach to categorize them into distressed and non-distressed groups. Nineteen financial variables are utilized to predict financial distress. Python is used as the programming language, and Jupyter Notebook facilitates algorithm development. The findings reveal that the LSTM model, when compared to RNN and CNN, achieved 91% accuracy using parameters such as 8 LSTM units with tanh activation and a dense layer with sigmoid activation function, a batch size of 10, 50 epochs, RMSprop optimizer, and binary cross-entropy loss were used. The study suggests that deep learning presents a novel approach that can enhance performance in financial distress prediction studies. This study is believed to be the first to utilize deep learning models for financial distress prediction in India based on single-year data, offering valuable insights for financial institutions and investors seeking more effective risk management strategies. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Real-Time Stress Monitoring Using IoT Wearable Sensors and Machine Learning
This research explores the potential of Internet of Things (IoT)-enabled wearable sensors in conjunction with machine learning techniques for real-time, non-invasive stress monitoring in women, using physiological data indicative of stress states. An IoT Wearables Dataset for Womens Safety: Stress Detection and Analysis was sourced from IEEE Data port and was used to train four supervised machine learning models, namely, random forest, support vector machines (SVM), gradient boosting, and logistic regression to classify individuals into categories of stressed and unstressed based on a predefined threshold of 0.35. The random forest algorithm attained the highest accuracy of 88.5% in categorizing stress, demonstrating reliable capabilities in identifying stress indicators from wearable sensor data. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Real-Time Stress Monitoring Using IoT Wearable Sensors and Machine Learning
This research explores the potential of Internet of Things (IoT)-enabled wearable sensors in conjunction with machine learning techniques for real-time, non-invasive stress monitoring in women, using physiological data indicative of stress states. An IoT Wearables Dataset for Womens Safety: Stress Detection and Analysis was sourced from IEEE Data port and was used to train four supervised machine learning models, namely, random forest, support vector machines (SVM), gradient boosting, and logistic regression to classify individuals into categories of stressed and unstressed based on a predefined threshold of 0.35. The random forest algorithm attained the highest accuracy of 88.5% in categorizing stress, demonstrating reliable capabilities in identifying stress indicators from wearable sensor data. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Bone Abnormality Detection Using RMSprop Optimizer in VGG16
The advent of deep learning has revolutionized medical imaging, enhancing diagnostic precision, treatment planning, and patient care. This study leverages deep learning, specifically employing the VGG16 model optimized with RMSprop, to automate bone abnormality detection. Methodologically, the research encompasses data acquisition, preprocessing, and model training with RMSprop optimization. Results highlight the efficacy of this approach, showcasing RMSprops ability to detect various bone abnormalities. These findings underscore deep learnings potential in medical imaging, emphasizing its applicability beyond bone abnormality detection. The study illuminates the transformative impact of RMSprop-optimized deep learning models in medical imaging, promising advancements in automated diagnosis and treatment planning. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Victimizing Cyberbullying Mental Illness Through Social Media
With the exponential increase of technologies and the growth of social media users, bullying takes different methods to reach its targets. Cyberbullying has been emerging lately in the form of bullying through voice memos, videos, and most frequently in the form of text messages. Bullies might use the rich and expansive environment that social networks offer to target their victims with their attacks. Cyberbullying has an especially negative impact on younger generations, who value social affirmation above everything. Studies have found a clear link between cyberbullying and suicidal ideation, particularly in teens. This concerning trend needs the development of efficient ways for identifying and combating cyberbullying, thereby protecting young lives. Many techniques can be used to identify the bullies linguistic patterns and create a detection model that will automatically identify instances of cyberbullying and whether it can lead to the happening of suicide or not. This project proposes a hybrid model of BiLSTM and EmoBERTa for detecting cyberbullying and checking the possibilities of it leading to the happenings of suicide. The dataset was run on different models and the proposed model yields the best average performance. By putting such detecting mechanisms in place, we can make the Internet safer. Early detection of cyberbullying enables intervention, which protects vulnerable people and may avoid disasters. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025. -
Machine Learning for Mental Health: A Sentiment Analysis Approach for Detecting Depressive Tendencies on LinkedIn During Layoffs Using RoBERTa
In the present corporate set-up, layoffs are an unfortunate yet common occurrence. Such occurrences lead to loss of job security and can have direconsequences on an individual's mental health, leading to depression. Depression was a global health concern well before the current downsizing came into the picture. These trying times have acted as a catalyst for this illness that affects not just mental health but all aspects of an individuals life. The study investigates the use of sentiment analysis on LinkedIn data to identify and examine depressive tendencies among victims of layoffs. Web-scraped information was taken from LinkedIn profiles of individuals affected directly or indirectly by layoffs. RoBERTa, a transformers model, is used to classify people as depressed or not by evaluating sentiment and emotional cues. A comparison between four machine learning algorithms- Decision Tree, Logistic Regression, SVM, and Nae Bayes is drawn to check their ability to detect depression. The SVM classifier performed best with an accuracy of 95.59% and 83.52% with the CountVectorizer and TF-IDF feature selection methods, respectively. Sentiment analysis aids in this research by examining the melancholic undertones in the words and phrases used in texts authored by people affected by layoffs directly or indirectly. The knowledge gained from this research can significantly affect corporate initiatives, mental health services, and human resource practices during such challenging times. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.