Browse Items (16481 total)
Sort by:
-
Defect originated photoluminescence tuning of silica nanoparticles prepared by electron beam irradiation and their applications
Considering the imminent importance of Silica (SiO2) nanoparticles (NPs), a highly rapid and one-pot scalable approach is being reported for their preparation. Electron-beam was used to derive the formation of SiO2 NPs, while in situ functionalization was carried out by ?-Cyclodextrin (?-CD). XRD pattern of as prepared ?-CD functionalized SiO2 NPs (i.e., ?-CD@SiO2 NPs) revealed their amorphous nature, while imaging studies showed self-assembling of NPs into a porous structure. UVvisible absorption spectra showed multiple peaks at 233, 323, 390 and 455 nm, which signifies the presence of different kind of defects in the as prepared NPs. An interesting aspect of this work is tuning of the photoluminescent properties of NPs from blue to green by simply varying the absorbed dose. This could be attributed to the formation of a particular kind of defects at a proportionate absorbed dose. These defects act as emission centers (ECs) and were analysed through steady state and time-resolved emission studies. Notably, ?-CD played significant role in influencing the composition of the NPs, whilst enhancing their colloidal stability and quantum yield. The prospective applications of ?-CD@SiO2 NPs were explored in latent fingerprinting and thermosensing. 2020 Elsevier Ltd and Techna Group S.r.l. -
Defect engineered unzipped multiwalled carbon nanotube/vanadium pentoxide composite for high-performance supercapacitor application
In the pursuit of next-generation energy storage systems, the advancement of high-performance electrode materials with enhanced capacitance and durability remains critical. This study presents a binary composite of unzipped multi-walled carbon nanotubes (UzMWCNTs) integrated with vanadium pentoxide (V2O5). The unzipping process introduces surface defects and oxygen functional groups, which enhance dispersion and provide numerous active sites. V2O5 nanoparticles uniformly anchor onto the UzMWCNT surface, offering pseudocapacitive behavior and boosting redox activity. The synergistic interaction between electric double-layer capacitance and faradaic charge storage delivers superior electrochemical performance. Structural and morphological characterization confirms successful composite formation, while electrochemical evaluations reveal a specific capacitance of 1135 F g?1 and cycling stability with 88% retention over 2000 cycles. This work highlights the potential of UzMWCNT/V2O5 hybrids as promising candidates for high-efficiency, next-generation supercapacitor electrodes. This journal is The Royal Society of Chemistry, 2026 -
Defect Engineered Few Layered MoS2 for HumanMachine Interface
Ultrasensitive flexible devices have huge applications in many areas, like healthcare monitoring, humanmachine interaction, and wearable technology. However, improving the sensitivity of these devices is still challenging. In the current study, a flexible non-contact sensing system is designed with a humanmachine interface using defect-engineered, few-layered Molybdenum disulfide (MoS2). The fabricated sensors show high sensitivity when monitoring proximity, humidity, and in-plane applied strain. The output performance demonstrates the influence of surface defects, which greatly impact the average surface charge of the nanosheets. The experimental measurements and in-detail density functional theoretical (DFT) calculation further reveal surface charge variations on the basal planes that correlate with topographic defects and increase sensitivity. The electrical signals for different gestures of human hands are used to illustrate the identification of multidirectional bending and sliding events. These findings will contribute to understanding the effect of surface defects that play an important role in sensing applications with humanmachine interface. 2025 Wiley-VCH GmbH. -
DeepRetina: Transformer-Enhanced EfficientNet for Retinal Disease Classification
Retinal diseases are a major cause of visual impairment in India, which requires precise and automated diagnosis tools.This paper, introduce a two-phase deep learning architecture for classifying five common retinal ailments: Glaucoma, Normal Fundus, Pathological Myopia, Hypertensive Retinopathy, and Cataract. A Swin Transformer (Swin-T) was fine-tuned on augmented retinal fundus images in the first phase to extract domain-adapted feature representations. The transformer utilize such embeddings for learning a regularized EfficientNet-inspired classifier in the second phase, with mixup augmentation and label smoothing for improving generalizability. Comprehensive experiments conducted on a carefully curated dataset of 643 test images validate that our method attains a test accuracy of 93.93%, with high precision as well as recall across all categories. The suggested pipeline strikes a suitable balance between feature abundance with transformer-based adaptation and resilient classification with EfficientNet, providing a viable tool for automated diagnosis of retinal ailments in practical clinical scenarios. 2026 IEEE. -
DeepBBBP: High Accuracy Blood-brain-barrier Permeability Prediction with a Mixed Deep Learning Model
Blood-brain-barrier permeability (BBBP) is an important property that is used to establish the drug-likeness of a molecule, as it establishes whether the molecule can cross the BBB when desired. It also eliminates those molecules which are not supposed to cross the barrier, as doing so would lead to toxicity. BBBP can be measured in vivo, in vitro or in silico. With the advent and subsequent rise of in silico methods for virtual drug screening, quite a bit of work has been done to predict this feature using statistical machine learning (ML) and deep learning (DL) based methods. In this work a mixed DL-based model, consisting of a Multi-layer Perceptron (MLP) and Convolutional Neural Network layers, has been paired with Mol2vec. Mol2vec is a convenient and unsupervised machine learning technique which produces high-dimensional vector representations of molecules and its molecular substructures. These succinct vector representations are utilized as inputs to the mixed DL model that is used for BBBP predictions. Several well-known benchmarks incorporating BBBP data have been used for supervised training and prediction by our mixed DL model which demonstrates superior results when compared to existing ML and DL techniques used for predicting BBBP. 2022 Wiley-VCH GmbH. -
Deep-fake Detection for Recognising Altered Audio using Deep Learning Approach
Ensuring the validity of audio recordings is becoming increasingly difficult due to deep-fake technology. Audio-analysis is used to identify deep-fake audio, which has been examined here. Machine-learning models can be made technological to compare between real and modified audio by examining minute artifacts and inconsistencies added to during the deep-fake production process. In this work, advanced signal-processing techniques like spectrum-analysis, voice-activity detection, and speaker-recognition; are used to extract relevant information from audio recordings. In order to exact deep-fake audio detection, these features are then utilized to guide and judge deep-learning models, such as recurrent neural networks (RNNs) and convolutional neural networks (CNNs). The objective is to create reliable and efficient techniques for detecting altered audio, almost eliminating the possible dangers. The goal is to provide reliable and efficient techniques for detecting modified audio in order to mitigate the possible risks related to deep-fake technology in a number of fields, such as social-media, journalism, and security. 2025 IEEE. -
Deep vs. Shallow: A Comparative Study of Machine Learning and Deep Learning Approaches for Fake Health News Detection
Internet explosion and penetration have amplified the fake news problem that existed even before Internet penetration. This becomes more of a concern, if the news is health-related. To address this issue, this research proposes Content Based Models (CBM) and Feature Based Models (FBM). The difference between the two models lies in the input provided. The CBM only takes news content as the input, whereas the FBM along with the content also takes two readability features as the input. Under each category, the performance of five traditional machine learning techniques: - Decision Tree, Random Forest, Support Vector Machine, AdaBoost-Decision Tree and AdaBoost-Random Forest is compared with two hybrid Deep Learning approaches, namely CNN-LSTM and CNN-BiLSTM. The Fake News Healthcare dataset comprising 9581 articles was utilized for the study. Easy Data Augmentation technique is used to balance this highly imbalanced dataset. The experimental results demonstrate that Feature Based Models perform better than Content Based Models. Among the proposed FBM, the Hybrid CNN - LSTM model had a F1 score of 97.09% and AdaBoost-Random Forest had a F1 Score of 98.9%. Thus, Adaboost-Random Forest under FBM is the best-performing model for the classification of fake news. 2013 IEEE. -
Deep Reinforcement Learning with Meta-Learning and Signal Bands for Indian Equity Portfolio Management
The portfolio is a collection of assets belonging to an investor. Managing the portfolio depends on the goal of the portfolio management. This paper proposed a new portfolio managing technique using a deep reinforcement learning framework combined with meta-learning and signal bands to optimize the returns and risk of the Nifty 50 index. The objective is to maximize portfolio returns by minimizing the risk, portfolio volatility, and drawdowns with constraints of transaction cost, maximum and minimum allocation, and availability of cash and holdings. The model executes the actions of buy, sell, and hold with the constraints, and the model executes any of those actions depending on the situation and model training. Proposed model recorded a 4.68 Sharpe ratio and 7.53 Sortino ratio while training the model. While testing the model, it recorded a 4.5 Sharpe ratio and 7.64 Sortino ratio, which aligns with the aim to achieve a higher Sortino and Sharpe ratio to build a robust model for risk-adjusted returns. Proposed approach aims to create a strong model for a portfolio management system that adapts to dynamic market conditions and optimizes investment strategies by integrating these techniques. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026. -
Deep Reinforcement Learning for Dynamic Resource Allocation in Intelligent Communication Networks
As intelligent communication implementations like 5G and IoT-enabled infrastructure meet technological advancement, provisioned network resource allocation dynamically and efficiently will be deemed crucial to accommodate diverse service demands and guarantee an optimized part of the network on demand. Resource management strategies based on traditional approaches often lack a sufficient response to the dynamic posture of network states and complex, heterogeneous environments. To address these challenges, deep reinforcement learning (DRL) has emerged as a powerful methodology wherein deep neural networks are employed to enable intelligent and adaptive decision-making based on dynamic network conditions. In this paper, we study the potential of DRL for dynamic resource management in intelligent communication networks. We build a DRL-driven agent that enables optimal allocation policy learning by interacting with high-dimensional, stochastic network environments with variable traffic loads, user mobility, and heterogeneous quality-of-service (QoS) requirements. Realistic simulation scenarios show that the proposed DRL framework outperforms conventional allocation heuristics in terms of throughput, latency, and fairness among users. We elaborate on the ramifications of explorationexploitation tradeoffs, convergence stability, and compute efficiency in the context of scale deployments. This way, our results prove that DRL is a potential candidate for dynamic resource allocation in future intelligent communication networks due to its better adaptability and performance. 2025 IEEE. -
Deep Reinforcement Learning and Its Industrial Use Cases: AI for Real-World Applications
This book serves as a bridge connecting the theoretical foundations of DRL with practical, actionable insights for implementing these technologies in a variety of industrial contexts, making it a valuable resource for professionals and enthusiasts at the forefront of technological innovation. Deep Reinforcement Learning (DRL) represents one of the most dynamic and impactful areas of research and development in the field of artificial intelligence. Bridging the gap between decision-making theory and powerful deep learning models, DRL has evolved from academic curiosity to a cornerstone technology driving innovation across numerous industries. Its core premiseenabling machines to learn optimal actions within complex environments through trial and errorhas broad implications, from automating intricate decision processes to optimizing operations that were previously beyond the reach of traditional AI techniques. Deep Reinforcement Learning and Its Industrial Use Cases: AI for Real-World Applications is an essential guide for anyone eager to understand the nexus between cutting-edge artificial intelligence techniques and practical industrial applications. This book not only demystifies the complex theory behind deep reinforcement learning (DRL) but also provides a clear roadmap for implementing these advanced algorithms in a variety of industries to solve real-world problems. Through a careful blend of theoretical foundations, practical insights, and diverse case studies, the book offers a comprehensive look into how DRL is revolutionizing fields such as finance, healthcare, manufacturing, and more, by optimizing decisions in dynamic and uncertain environments. This book distills years of research and practical experience into accessible and actionable knowledge. Whether you're an AI professional seeking to expand your toolkit, a business leader aiming to leverage AI for competitive advantage, or a student or academic researching the latest in AI applications, this book provides valuable insights and guidance. Beyond just exploring the successes of DRL, it critically examines challenges, pitfalls, and ethical considerations, preparing readers to not only implement DRL solutions but to do so responsibly and effectively. Audience The book will be read by researchers, postgraduate students, and industry engineers in machine learning and artificial intelligence, as well as those in business and industry seeking to understand how DRL can be applied to solve complex industry-specific challenges and improve operational efficiency. 2024 Scrivener Publishing LLC. -
Deep Q-Learning for Autonomous Vehicle Navigation in Smart Mobility
The proposed system leverages Deep Q-Learning to enhance autonomous vehicle navigation in smart mobility environments. By integrating reinforcement learning with deep neural networks, the system enables vehicles to make real-time decisions while adapting to dynamic traffic conditions. The framework employs a reward-based learning mechanism to optimize path selection, collision avoidance, and efficient maneuvering in complex urban scenarios. To improve decisionmaking accuracy, the proposed approach incorporates an experience replay mechanism, preventing overfitting and ensuring stable learning. Additionally, a target network is utilized to enhance training convergence, allowing the model to generalize effectively across varying road conditions. The system is further optimized through adaptive explorationexploitation strategies, enabling vehicles to balance learning new routes while prioritizing safe and efficient navigation. The proposed methodology demonstrates significant improvements in autonomous mobility, offering a scalable and robust solution for next-generation smart transportation systems. 2025 IEEE. -
Deep neural network architecture and applications in healthcare
Gaining insights related to medical data has always been a challenge, as limited technology delays treatment. Various types of data are collected from the medical field, such as sensor data, that are heterogeneous in nature. All of these are very poorly maintained and require more structuring. For this reason, deep learning is becoming more and more popular in this area. There are many challenges due to inadequate and irrelevant data. Insufficient domain knowledge also adds to the challenge. Modern deep learning models can help understand the dataset. This chapter provides an overview of deep learning, its various architectures, and convolutional neural networks. It also highlights how deep learning technologies can help advance healthcare. 2022 River Publishers. -
Deep learning: Research and applications
This book focuses on the fundamentals of deep learning along with reporting on the current state-of-art research on deep learning. In addition, it provides an insight of deep neural networks in action with illustrative coding examples. Deep learning is a new area of machine learning research which has been introduced with the objective of moving ML closer to one of its original goals, i.e. artificial intelligence. Deep learning was developed as an ML approach to deal with complex input-output mappings. While traditional methods successfully solve problems where final value is a simple function of input data, deep learning techniques are able to capture composite relations between non-immediately related fields, for example between air pressure recordings and English words, millions of pixels and textual description, brand-related news and future stock prices and almost all real world problems. Deep learning is a class of nature inspired machine learning algorithms that uses a cascade of multiple layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. The learning may be supervised (e.g. classification) and/or unsupervised (e.g. pattern analysis) manners. These algorithms learn multiple levels of representations that correspond to different levels of abstraction by resorting to some form of gradient descent for training via backpropagation. Layers that have been used in deep learning include hidden layers of an artificial neural network and sets of propositional formulas. They may also include latent variables organized layer-wise in deep generative models such as the nodes in deep belief networks and deep boltzmann machines. Deep learning is part of state-of-the-art systems in various disciplines, particularly computer vision, automatic speech recognition (ASR) and human action recognition. Tutorials on deep learning framework with focus on tensor flow, keras etc. Numerous worked out examples on real life applications Illustrative diagrams and coding examples. 2020 Walter de Gruyter GmbH, Berlin/Boston. All rights reserved. -
Deep learning-driven correction of motion-induced artifacts in microfluidic on-chip fluorescence microscopy for robust cell classification
Fluorescence microscopy combined with microfluidic platforms allows for the analysis of single cells and the whole biomedical process to be done at high speed, however, it is often a very delicate method that can be heavily affected by motion-induced distortions during the high-speed flow. These artifacts, such as motion blur, misalignment, and shape deformation significantly lower automatical accuracy of the cell classification. The suggested research suggests that on-chip fluorescence microscopy employs an AI-based framework of distortion correction using Vision Transformers (ViT) and Generative Adversarial Networks (GAN) to remove motion artifacts in real-time. The combination of the GAN-ViT architecture does not only manage to reconstruct image quality but also to preserve fine cellular features when flowing system rates increase to 200 4L/min, which provide PSNR = 38.6 dB and SSIM = 0.98. When the system was used in both synthetic and experimental microfluidic data, it was able to reach a classification accuracy of 99.9, thereby indicating consistency in the system despite varying flow rates. The speed of the framework is 950 frames per second (fps), almost equal to the 1000-fps smartphone camera acquisition rate, thereby, demonstrating its suitability to the real-time, high-throughput imaging. As opposed to the past CNN or transformer techniques, a hybrid GAN-ViT architecture offered by the authors of this study directly implements in the imaging pipeline, thus enabling the simultaneous motion correction and diagnostic classification to occur immediately. The study results highlight the fact that AI-based distortion correction not only increases the accuracy of the diagnosis, but also personnel and laboratory response in microfluidic fluorescence microscopy. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2026. -
Deep Learning-Based Signal Detection Techniques for Real-Time Communication in Fading Channels
Dependable signal detection has also been a major concern in real-time wireless communication especially in the case of fading channels that cause non-adaptive distortion and deteriorate the overall performance drastically. The conventional detection meth-ods, like the maximum likelihood detection, are not always adaptive in the circumstances of dynamic and therefore unpredictable channel conditions, and particularly in the cases when the statistical profiles are unknown or vary too quickly. In order to address these shortcomings, the papers introduce a new paradigm of deep learning signal detection trained to learn hierarchies and temporal patterns of raw received signals, which by their pas integrate convolutional neural networks (CNN) and recurrent neural networks (RNN). The trained architecture is end-to-end that is able to map the noisy distorted inputs to their symbols which are inherently transmitted in the context of channel state informa-tion. Heavy simulation over Rayleigh and Rician fading channels with different Doppler spreads and SNR values shows that the suggested approach shows substantial improve-ment over the traditional maximum likelihood and classical machine learning-based detec-tors regarding bit error rate (BER), inference latency and computational overhead. Such results emphasize the performance as well as the flexibility of deep learning model in very dynamic propagation conditions. On the whole, this paper draws the conclusion that deep learning is a perspective direction to solve the problem of real-time detection of a signal in next-generation wireless networks, such as a 6G or IoT edge setup. 2025, Society for Communication and Computer Technologies. All rights reserved. -
Deep Learning-Based Prediction of Physical Activity Intensity for Athletes
Maximizing training plans for athletes and lowering the risk of injury depends on a precise assessment of the degree of physical activity. Existing system in-use systems often employ simplistic models, which leads to inaccurate projections. The paper presents a deep learning-based system that uses convolutional neural networks (CNNs) to create real-time predictions using wearable sensor data. Because it automatically extracts relevant features from raw sensor data, the technique does not need human feature engineering. Utilizing thorough model training and evaluation, it exceeded the most recent methods in terms of accuracy (0.92), precision (0.90), recall (0.92), F1-score (0.91), and ROC AUC (0.94). Results of cross-validation over many data subsets confirm the resilience of the method. Comparisons of confusion matrices also demonstrate how effectively the algorithm forecasts various activity intensities. Overall, the proposed system represents a breakthrough in accurately estimating how much physical activity athletes do, enhancing the efficacy of their training, and reducing the possibility of damage in sporting settings. 2024 IEEE. -
Deep Learning-Based Optimised CNN Model for Early Detection and Classification of Potato Leaf Disease
After rice and wheat, potatoes are the third-largest crop grown for human use worldwide. The different illnesses that can harm a potato plant and lower the quality and quantity of the yield cause potato growers to suffer significant financial losses every year. While determining the presence of illnesses in potato plants, consider the state of the leaves. Early blight and late blight are two prevalent illnesses. A certain fungus causes early blight, while a specific bacterium causes late blight. Farmers can avoid waste and financial loss if they can identify these diseases early and treat them successfully. Three different types of data were used in this study's identification technique: healthy leaves, early blight, and late blight. In this study, I created a convolutional neural network (CNN) architecture-based system that employs deep learning to categorise the two illnesses in potato plants based on leaf conditions. The results of this experiment demonstrate that CNN outperforms every task currently being performed in the potato processing facility, which needed 32 batch sizes and 50 epochs to obtain an accuracy of about 98%. 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Deep Learning-Based Health Risk Prediction in Contact Sports Using Wearable Sensor Data
This study presents a deep learning-based approach to predicting physiological health risks in athletes engaged in contact sports using wearable sensor data. Motivated by the need to detect early warning signs of collapse or severe fatigue, this study employs a Long Short-Term Memory (LSTM) neural network to analyse multivariate time-series data. Key physiological signals, including heart rate, body temperature, and motion, were extracted from the PAMAP2 dataset to train and validate the model. The LSTM demonstrated strong predictive performance, achieving an accuracy of 98.3% in identifying potentially dangerous physiological states. In addition to its high classification accuracy, the model effectively captured temporal dependencies in the data, underscoring its suitability for health risk prediction in dynamic, high-intensity sports environments. This study highlights the potential of wearable data and LSTMbased analysis in supporting proactive athlete health management and injury prevention. 2025 IEEE. -
Deep Learning-based Gender Recognition Using Fusion of Texture Features from Gait Silhouettes
The gait of a person is the manner in which he or she walks. The human gait can be considered as a useful behavioral type of biometric that could be utilized for identifying people. Gait can also be used to identify a persons gender and age group. Recent breakthroughs in image processing and artificial intelligence have made it feasible to extract data from photographs and videos for various classifying purposes. Gender can be regarded as soft biometric that could be useful in video captured using surveillance cameras, particularly in uncontrolled environments with erratic placements. Gender recognition in security, particularly in surveillance systems, is becoming increasingly popular. Popularly used deep learning algorithms for images, convolutional neural networks, have proven to be a good mechanism for gender recognition. Still, there are drawbacks to convolutional neural network approaches, like a very complex network model, comparatively larger training time and highly expensive in computational resources, meager convergence quickness, overfitting of the network, and accuracy that may need improvement. As a result, this paper proposes a texture-based deep learning-based gender recognition system. The gait energy image, that is created by adding silhouettes received from a portion of the video which portrays an entire gait cycle, can be the most often utilized feature in gait-based categorization. More texture features, such as histogram of oriented gradient (HOG) and entropy for gender identification, have been examined in the proposed work. The accuracy of gender classification using whole body image, upper body image, and lower body image is compared in this research. Combining texture features is more accurate than looking at each texture feature separately, according to studies. Furthermore, full body gait images are more precise than partial body gait images. 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. -
Deep Learning-Based Dynamic Vision: Classifying Hand Gestures
In the field of hand gesture recognition, this research introduces novel approaches by utilising a variety of state-of-the-art deep learning models, including YOLOv6, YOLOv8, VGG16, VGG19, and ResNet50. Our work involved rigorous dataset annotation and preprocessing, coupled with custom data augmentation techniques tailored for real-world scenarios. The results were excellent, as YOLOv6 exhibited remarkable precision, achieving an impressive Average Precision (AP) of 97.4% and recall (AR) of 90%. Meanwhile, YOLOv8s prowess shone in specific classes, where it attained a remarkable mean Average Precision (mAP) of 89%. We further explored the capabilities of classical Convolutional Neural Networks (CNNs) such as VGG16 and VGG19. These models demonstrated solid performance with an average accuracy of 74 and 67%, respectively. Our study also explored the utilization of ResNet50, which, despite its popularity in other computer vision tasks, showed a lower accuracy of 33% in the context of hand gesture recognition. This research showcases a significant leap beyond the conventional CNN-based research in hand gesture recognition, as we integrated both object detection and image classification models into the evaluation framework. Looking ahead, our research opens doors to exploring ensemble models that synergize the strengths of YOLOv6, YOLOv8, VGG16, and VGG19, promising a harmonized performance across all classes. Moreover, we advocate for further research into transfer learning techniques, anticipating even higher accuracy levels in scenarios constrained by limited training data. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.