Proceedings of 6th International Conference on Recent Trends in Machine Learning, IoT, Smart Cities and Applications

Inhaltsverzeichnis

1. Frontmatter

2. CNN-XGBoost with Score-CAM for Detection of Cardiovascular Diseases Using ECG Images

   Pillai Lekshmi Ashokan, S. Siva Sathya

   Abstract

   The heart, being a vital organ in the human body, requires continuous monitoring to detect various cardiovascular diseases. Electrocardiography (ECG) has been one of the oldest and most effective methods for diagnosing heart diseases, providing critical insights into the heart’s electrical activity. However, interpreting ECG signals is complex, requiring trained and expert cardiologists, leading to potential delays in diagnosis and treatment. To address this challenge, an automated diagnostic system is essential to facilitate timely and accurate detection of heart conditions. In this research, we introduce a hybrid model combining Convolutional Neural Networks (CNN) and XGBoost as one whole architecture to classify ECG images into four categories: Normal, Myocardial Infarction (MI), Abnormal, and Previous History of MI. Our model achieves an impressive accuracy of 98%, along with excellent precision, recall, and F1 scores, demonstrating its robustness in handling multi-class classification. The need for transparency in the machine and deep learning models, particularly in healthcare, is crucial for clinical acceptance. To ensure interpretability, we employ the Score-CAM (Score- Weighted Class Activation Maps) technique, which generates saliency maps to visualize the important regions of ECG images that influence the model’s decision-making process. These visualizations provide valuable insights into the model’s focus on diagnostically significant features such as the QRS complex, ST-segment elevation, and T wave patterns, thereby enhancing the trust and reliability of the model in medical applications.

3. Personalized Depression Management System Using LLMs and Reinforcement Learning: A Survey

   Vaishali Katti, Kailash J. Karande

   Abstract

   Depression, a complex and pervasive mental health disorder, affects millions globally and demands personalized, adaptive treatment approaches. Recent advancements in artificial intelligence, particularly the use of Large Language Models (LLMs) and Reinforcement Learning (RL), offer transformative opportunities for digital mental health support. This survey explores the intersection of LLMs and RL in developing personalized depression management systems. LLMs, such as GPT-based models, excel at generating human-like, empathetic dialogue, enabling conversational agents to simulate therapeutic interactions. These models can respond to user sentiment, detect linguistic patterns related to depression, and deliver content aligned with cognitive behavioral therapy (CBT) principles. Meanwhile, RL algorithms focus on optimizing long-term therapeutic outcomes by learning from user interactions and adapting strategies based on feedback and reward mechanisms. They enable adaptive scheduling of interventions, behavioral nudging, and habit reinforcement. This paper reviews recent advancements, architectures, clinical applications, challenges, and comparative benefits of both technologies in the context of mental health. We highlighted existing systems, emerging hybrid models, and open research questions related to privacy, personalization, and ethical deployment. The survey concludes by outlining the potential of integrated LLM-RL frameworks to revolutionize depression care through scalable, dynamic, and emotionally intelligent support systems tailored to individual needs.

4. IoT-Enabled Smart Traffic Signal System for Ambulance Prioritization and Congestion Management

   Guna Santhoshi, Sai Varshitha Bellamkonda, Priyanka Kollu, K. Lakshmi Sanjana

   Abstract

   The “Smart Traffic Control with Ambulance Detection” technology prioritizes the ambulances at the traffic lights in response to an increase in the number of vehicles. There are red, green, and orange LEDs installed for the control of traffic on each of the four roads at the intersection. The infrared sensors are used to determine the density of each road, thereby allowing the system to dynamically change to green for the road with the highest density, thus relieving the traffic. It identifies the RFID tags placed on ambulances to give them priority on the road to ensure uninterrupted travel. The system combines manual police interventions with computerized traffic control whenever necessary. This project aims at enhancing the emergency response time and efficiency of traffic management by integrating RFID-enabled ambulance detection with density-based traffic control.

5. AI-Powered Mental Health Chatbot Using NLP and Sentiment Analysis

   A. Nageswari, Vidhi R. Shah, Shabana, Asian Kumari

   Abstract

   This paper presents the development of an AI-powered mental health chatbot that leverages Natural Language Processing (NLP) and sentiment analysis to assist users experiencing stress, anxiety, or depression. The chatbot employs an LSTM-based neural network for contextual understanding and VADER for sentiment polarity detection, enabling emotion-aware responses in real-time. Using custom datasets and fine-tuned models, the chatbot offers private and scalable mental health support, particularly beneficial in underserved regions.

6. Systematic Review and Primary Outcomes in Arrhythmia Detection via Evolutionary Optimization

   Md. Shamshad Begum, Nimmagadda Padmaja

   Abstract

   The accurate detection and classification of arrhythmias from electrocardiogram (ECG) signals is a critical challenge in clinical cardiology, with current models having limitations in sensitivity, computational efficiency, and generalizability. This study introduces the Human Evolutionary Liquid-Neural Network (HEL-Net), an advanced deep learning model that addresses these issues by combining the Human Evolutionary Optimization Algorithm (HEOA) and Liquid Neural Networks (LNN). HEL-Net uses HEOA to optimize feature selection, ensuring that only the most relevant features are used, thereby increasing the model’s efficiency and accuracy. Meanwhile, LNN processes dynamic temporal sequences, capturing the subtle variations found in arrhythmia signals. Initial results show that the model has the potential to provide high sensitivity and specificity in arrhythmia detection, and this is supported by a comprehensive methodological approach to feature optimization and classification. The paper reviews recent advancements, identifies gaps in current technologies, and presents preliminary findings that demonstrate HEL-Net’s effectiveness in clinical settings.

7. Network Intrusion Detection in VANETS Using Machine Learning: Securing VANETS with Machine Learning- Based NIDS Approach

   I. Ravi Prakash Reddy, Akshaya Juluri, Sanjana Reddy, Pola Rishika

   Abstract

   In response to the growing cybersecurity challenges in Vehicular Ad Hoc Networks (VANETs), we designed a Network Intrusion Detection System (NIDS) based on machine learning that can better detect malicious threats. The system is based on three main algorithms: Random Forest, XGBoost, Gradient Boosting. These algorithms classify network traffic as normal or malicious. By combining these algorithms, the system exploits their collective strengths, thereby improving the detection accuracy. The NIDS has an alert function that sends notifications about possible intrusions directly to the user’s web interface. This approach will improve road safety and traffic management by strengthening vehicular communication security. The system is trained and evaluated using the NLS-KDD dataset for better detection performance in VANET environments.

8. Combatting Malnutrition Through Personalized Web-Based Platforms: An Integrated BMI, BMR and TDEE Approach

   S. Ramacharan, Gummireddy Saranya, Nandula Vyjayanti, Gummadavelli Vaishnavi, Banda Sravanthi Yadav

   Abstract

   Malnutrition is a pressing issue in rural India that is caused by the lack of nutrients or by getting too many nutrients. The UN’s Goal “Zero Hunger” aims to reduce malnutrition. To combat malnutrition, we designed this user-friendly web platform. Initially, the system focused only on children under five; we now extended our platform services to the age group of 45 years. We enhanced our features to recommend season-specific recipes to improve nutritional adequacy. More important are the expanding uses of BMR and TDEE calculations in combination with BMI. BMR enables one to understand his or her rate of metabolism, which varies from person to person. This means that two individuals with the same BMI have different caloric needs depending on the BMR. By including both BMR and TDEE, we can, thus, provide more personalized nutrition information that is keyed into the individual’s metabolic health concerning their unique nutritional need and requirements.

9. Yield Direct: Helping Farmers Maximize Yield

   S. Ramacharan, Anvitha Moilla, K. Saba Nazneen, Geethanjali Rai

   Abstract

   Agriculture is the backbone of India’s economy, sustaining a large portion of the population. However, many farmers, particularly those managing small to medium-sized farms face substantial challenges in increasing crop yields and securing market access. YieldDirect is an innovative platform designed to address these issues by providing a comprehensive solution that integrates real-time data-driven crop management and direct-to-consumer sales capabilities. The platform leverages advanced machine learning techniques, such as Support Vector Machine [SVM], K-Nearest Neighbors [KNN], Convolutional Neural Networks [CNN], and Random Forest, to deliver personalized crop recommendations and disease predictions. Additionally, the ecommerce module enables farmers to bypass intermediaries, improving their profit margins by selling directly to consumers. Through YieldDirect, farmers can enhance productivity and access fairer markets, leading to more sustainable and profitable farming practices.

10. Smart Solar Grass Cutter with Lawn Coverage Assistant

    A. Vijaya Krishna, V. Apurva, Lingam Navya Sree, Patlolla Nikitha Reddy, C. Meghana

    Abstract

    The Smart Solar Grass Cutter with Lawn Coverage is an efficient and innovative way to maintain modern lawns while protecting the environment. This particular device uses solar power, making it less dependent on electricity or gasoline, reducing environmental impact and costs. As long as sunlight is not a problem, this device operates sustainably and thus is something ideal for eco-friendly homeowners. It leverages renewable energy sources to give efficient and cost-effective alternatives to traditional lawn care methods. It has sensors, equipped to recognize obstacles and comes to a safe halt if some obstacle is there in the passage. The motor comes with more robust overheat protection for it to increase durability. Precise as well as easy-cutting can be made from any position without much man labor because patterns are customizable through the application that works with smart connectivity. This innovative device combines sustainability, smart technology, and user-friendly design, promoting efficient, safe, and eco-conscious lawn care.

11. TOCyG-Net: A Triple-Integrated Model for Superior Rain Streak Removal in High-Fidelity Image Processing

    K. Hemavani, G. S. Annie Grace Vimala, G. Nalinipriya

    Abstract

    Recent advances in image processing have primarily focused on improving image quality when affected by environmental conditions such as rain. This paper introduces TOCyG-Net, a novel architecture that uses Total Generalized Variation (TGV), Cycle-Consistent Generative Adversarial Networks (CycleGAN), and Orthogonalized Iterative Shrinkage (OIS) to effectively remove rain streaks from images. The proposed method starts with TGV for initial image smoothing, which reduces noise while preserving important structural features like edges and texture. CycleGAN then facilitates the transition from a rain-impaired to a rain-free image state by learning the mapping between these domains. Finally, the OIS method refines the image further by reducing less significant coefficients, which improves overall image clarity and detail retention. TOCyG-Net outperforms existing models, with a Peak Signal-to-Noise Ratio (PSNR) of 41.26 and a Structural Similarity Index (SSIM) of 0.994, indicating superior detail preservation and noise reduction performance. This approach not only sets a new standard for rain streak removal, but it also paves the way for real-time image processing applications by addressing some of the common limitations of current techniques, such as high computational demands and training stability issues.

12. Advancing Media Integrity with VeritasBERT: A Specialized Approach to Fake News Identification

    T. V. Divya, Pavan Kumar Pagadala

    Abstract

    In this study, we created and tested a sophisticated fake news detection system using a proposed model called VeritasBERT, which combines Bidirectional Encoder Representations from Transformers (BERT) with a softmax classification layer. Our methodology included extensive data collection using web crawlers, preprocessing, and feature extraction in order to effectively train the model on a diverse dataset derived from various online sources, such as social media platforms and news sites. In comparison to traditional and existing models, the proposed model outperformed them on a variety of metrics. The VeritasBERT achieved 99.49% accuracy, 99.38% precision, 99.24% recall, and an F1 score of 99.45%. These results outperformed other tested models, including LSTM, NLP combined with LSTM, and hybrid models like Naive Bayes with Logistic Regression, with the highest accuracy of 99.2%, precision of 99.1%, and recall of 99%. Precision-recall curves demonstrated the model’s efficiency by confirming its high precision even at varying recall thresholds, making it extremely reliable for practical applications. Error analysis revealed that the model’s main challenges are distinguishing between ‘Suspicious’ and ‘Real’ news, which will guide future system improvements. This study demonstrates the effectiveness of using advanced deep learning techniques to address the complexities of fake news detection, providing robust solutions that can be applied to real-time news verification systems to improve information credibility across digital platforms.

13. Advanced Machine Learning for CAPTCHA Recognition: Evaluating the Efficacy of FlowCap-Net

    Kumar Akuthota, Saritha Anchuri, Yellaturu Deekshitha, R. Praveen Kumar Naidu, Bonthala Balaji, A. Basi Reddy

    Abstract

    This study explores advanced machine learning solutions for CAPTCHA recognition, a critical aspect of web security. We have developed a new model, FlowCap-Net, which leverages the strengths of Inception V4 and Liquid Neural Networks to effectively address the challenges of CAPTCHA recognition. FlowCap-Net stands out in our comprehensive evaluation, achieving an exceptional accuracy rate of 98.95%. This model demonstrates remarkable efficiency and precision in deciphering complex CAPTCHA designs, establishing itself as a potent tool against the evolving threats in digital security environments. Our findings highlight FlowCap-Net as a groundbreaking advancement in the field, offering substantial improvements over traditional CAPTCHA recognition technologies.

14. Hybrid GAN and ResNet-50 Model for Deepfake Detection Using AI-Powered Face-Fusion Analysis

    Karamala Naveen, Saritha Anchuri, Bachala Varshitha, R. Praveen Kumar Naidu, Motam Roopa Sree, A. Basi Reddy

    Abstract

    Deepfake detection is a critical area in digital media, given the rapid advancement and accessibility of deepfake technology. This research paper introduces a sophisticated AI-powered deepfake detection system that leverages the comprehensive FaceForensics++ dataset. Advanced machine learning algorithms including FaceNet for landmark detection, ResNet-50 for robust feature extraction, and GANs for synthetic face generation and detection are used in the system. The presentation of the exemplary is calculated using rigorous metrics, achieving an accuracy of 94.25%, precision of 93.33%, recall of 94.26%, and F1-score of 94.24%, with a mean squared error (MSE) of 0.1775, demonstrating substantial efficacy in identifying manipulated content, and an AUC of 0.99, indicating exceptional accuracy in distinguishing genuine and manipulated videos. To stay ahead of growing deepfake methods, the model will be improved to generalize to unseen alterations and detect in real time.

15. Structural Analysis of Irregular Structure with and Without Shear Wall and Bracing

    Sangram Mule, C. P. Pise, A. A. Kamble

    Abstract

    Earthquake is the one of the main natural disaster for human inhabitation. Disaster causes drastic damage to man-made structures like High rise building and Infrastructures. In this paper main focus is on the high-rise building. The high rise building classified as Regular shape building and Irregular shape building. While designing a irregular building several challenges occurs which can be not be neglected. Primarily concerning aspects of Irregular building are structural analysis, building planning, construction complexity and also building performance under different kind of loadings and design parameters.

    In this work provision of shear wall and bracing system is taken into consideration to enhance the lateral stiffness of building, Ductility and to take measures against the lateral displacement. Two type of building models are developed and analysis by E-tab software. In the present work G + 14 multi storey building is analyzed with using shear wall and bracing system. It is analyzed and results of storey displacement, Drift, Overturning Moment & Base shear are evaluated and compared with normal shear wall and bracing system model.

16. AI-Powered Virtual Fashion Assistant

    S. Radhika, A. Sathwika, A. Naga Bhavani, K. Rinda Adithi

    Abstract

    The Virtual Fashion Assistant is a cutting-edge application that redefines online shopping by providing personalized fashion recommendations and virtual try-ons using advanced computer vision, machine learning, and deep learning technologies. Tailored to each user’s body shape and style preferences, the system allows users to upload images of themselves and garments, generating realistic virtual try-on experiences that support informed, interactive clothing choices. Additionally, the system enhances fashion discovery by scraping online retailers for similar styles, offering a seamless and comprehensive shopping experience This blend of human pose estimation, body analysis, and web scraping enables users to confidently select clothing that fits well and matches their unique taste, positioning the Virtual Fashion Assistant as a pioneering solution in the field of virtual fashion and e commerce.

17. Integrated System for Leaf Disease Prediction and Crop Recommendation Using Machine Learning and IoT

    Shradha Jadhav, Swati Bhisikar, Pranali Jadhav, Pranjali Jadhav

    Abstract

    The convergence of the Internet of Things (IoT) and Machine Learning in agriculture has drastically been changing traditional agricultural practices toward an intelligent, efficient future with the use of data-driven machinery. This paper suggests an integration approach towards predicting leaf disease in real-time and crop recommendation based on machine learning techniques (ML) with the help of internet-of-things devices (IoT). Leaf disease prediction is important to detect plant diseases at an early stage to prevent them, thus affecting agriculture’s productivity. It uses IoT sensors for environment sensing and machine learning algorithms for crop recommendation based on data provided by sensors in real time. The study will concern small-scale farms, which rely on resource management and in-time detection of infections to increase crop yield. IoT and ML-related techniques are combined in the proposed solution to provide a full-stack framework through which farmers can make informed decisions for improved agricultural productivity.

18. Tweeting the Vote: Influence Propagation and Community Analysis on Political Tweets with a Quantum Clustering Approach

    Elizabeth Leah George, Subashini Parthasarathy

    Abstract

    Computational social science research on political tweets is a general study that examines political communication dynamics on social media during the April 2024 Indian elections, focusing on user perceptions of political tweets. A primary dataset of scraped tweets underwent preprocessing, including tokenisation, lemmatisation, and stop-word removal. Sentiment analysis using the NLTK package revealed that 51.75% of tweets were positive, 41.23% negative, and the rest expressed neutral sentiments with emotional undertones like fear, surprise, and sadness. A comprehensive evaluation used classical, ensemble-based, and quantum-inspired clustering and classification methodologies. Classical methods like KMeans showed moderate results (accuracy: 32.46%, modularity: 0.42, silhouette score: 0.35), improving significantly when paired with a Support Vector Classifier and a Radial Basis Function kernel (accuracy: 85.09%, modularity: 0.46). Ensemble methods improved clustering quality, with Bagging Ensemble achieving a modularity of 0.58, silhouette score of 0.47, and an F1 score of 58.22%. Quantum-inspired approaches showed notable performance, with Quantum K-Means using dynamic centroids initialisation achieving the highest modularity (0.57), silhouette score (0.48), and Calinski-Harabasz Index (545.12). Quantum-Enhanced DBSCAN delivered the best accuracy (70.02%) and F1 score (68.01%). Community detection algorithms like Louvain and Newman-Girvan further provided insights into thematic coherence. This study underscores the potential of advanced clustering methodologies for effectively analysing political discourse on social media.

19. Comparative Structural Analysis of Conventional and Pre-engineered Building

    Narendra M. Shete, C. P. Pise, Amol Kamble

    Abstract

    Recent advancements in the construction industry have influenced many aspects of human life, like in terms of economic feasibility, structural safety, and the fast pacing of construction activities. Historically, metal structures played a Important role during the early 20^th century, notably aiding in infrastructure development throughout the Industrial Revolution and the challenges of World War II. In the present-day structural engineering, Pre-Engineered Buildings (PEBs) have emerged as a modern and efficient alternative to Conventional Steel Buildings (CSBs), giving solution to several limitations that arises with traditional construction techniques.

    Typically, industrial buildings may be classified into two structural categories: (1) structures with short spans employing continuous or cantilevered girders over columns, and (2) longer-span steel-framed industrial buildings supported by columns. In the present study, a comparative analysis between PEB and CSB systems is conducted for an industrial warehouse using STAAD.Pro. Main indicators such as base shear, storey drift, and the influence on structural stability were observed. The objective was to determine the optimal structural configuration under varying seismic and wind loading cases. Both structural models were analyzed and designed according to Indian Standards, IS 1893 for seismic loading and IS 875 Part 3 (2015) for wind load data.

20. Design and Implementation of a Scalable Enterprise Integration Platform Using Microservices and Cloud Technologies

    K. M. Deepika, Ankur Keshri, Devank Gupta, Mohiuddin Hasan, Saishwar Anand, Adarsh Nagamangala Sreenivasan, Praveen Kumar Sharma

    Abstract

    This paper presents the design and implementation of a scalable Enterprise Integration Platform (EIP) built on a modern microservices architecture, utilizing technologies such as ReactJS, Node.js, MongoDB, and Redis. The platform addresses the challenges of integrating heterogeneous enterprise systems by supporting structured and unstructured data via SQL and NoSQL databases. It emphasizes real-time messaging, continuous integration/deployment (CI/CD), and hybrid cloud deployment, ensuring agility and operational efficiency. Performance evaluation demonstrates the platform’s ability to support high throughput and concurrent users, making it suitable for cloud-native, enterprise-grade integration scenarios.

21. Automated Research Paper Classification and Recommendation System for Efficient Literature Discovery

    Sree Lakshmi Done, Jananee Schandra, G. Rishitha, M. Jasmita Rani

    Abstract

    This paper proposes an AI-based system to classify research papers in the relevant academic domain and recommend equally similar papers based on the title and essence of a given paper. Using a semantic scholar and scholar API, the system automatically reduces the subject-field classification and relevance-based paper recommendations, leaving the necessary manual efforts in literature reviews. This system facilitates metadata such as quotes, author details and intensive explorations like direct paper URL. A user -friendly web interface enables spontaneous interactions, enhancing educational productivity by accelerating the discovery of literature and improving the search accuracy.

22. Performance Evaluation of Cost-Effective Standalone Solar Still System with Solar Concentrator

    Dongare Santosh, Sunil Somani, Hiralal Patil

    Abstract

    This research explores the design, development, and performance evaluation of a compact, standalone solar still system for efficient water purification. With the growing challenge of freshwater scarcity worldwide, solar stills offer a sustainable approach by harnessing solar energy for desalination and purification. The proposed system is designed to be compact, affordable, and easy to install, making it particularly suitable for remote and off-grid locations. The construction of the solar still focuses on optimizing its dimensions to enhance solar absorption and condensation efficiency. Key design elements include a high-transparency glass cover for improved sunlight penetration, a black-coated basin to maximize heat absorption, and a well-sealed structure to reduce heat loss. A condenser composed of thermally conductive materials is incorporated to enhance the speed of condensation. To evaluate its efficacy, performance tests were executed under diverse climatic conditions, quantifying variables such as water output, sun radiation, ambient temperature, and humidity. The results demonstrate that, under ideal conditions, the system may produce up to 6.7 l of cleaned water per square metre per day, with maximum efficiency occurring during periods of high solar intensity. This study highlights the potential of a compact, standalone solar still as a reliable and sustainable solution for water purification, especially in regions facing freshwater shortages and lacking electricity access.

23. Real Time Intrusion Detection System Using AI&ML

    Sangita M. Jaybhaye, Pratham P. Chintawar, Chetan D. Niwate, Aditya R. Narke, Arya V. Sawant, M. D. Jaybhaye

    Abstract

    Preventing cyber-attacks like DoS, Probe, R2I, and U2R intrusions in a digitally interconnected world is very essential. This paper proposes a Real-Time Intrusion Detection System (RIDS) that identifies known and unknown intrusions using AI and ML. The system uses supervised and unsupervised ML algorithms, along with advanced preprocessing and feature selection, to enhance real-time anomaly detection and minimize false positives. It combines conventional and deep learning techniques with the Random Forest (RF) methodology to achieve higher accuracy and improve scalability. Experimental evaluations conducted on standard datasets show the effectiveness of intrusion detection while simultaneously reducing computational requirements.

24. Smart Glove: A Hand Gesture Recognition System and Home Automation

    Sonali K. Godase, Anita H. Shinde, Anuja N. Patil, Sakshi V. Narayanpethkar, Gauri R. Kadam, Swapnil R. Takale

    Abstract

    Communication is essential for human connection, but individuals with speech impairments often struggle to express themselves effectively. While sign language serves as a powerful tool for communication, its impact is limited because many people do not understand it. To bridge this gap, we have developed a MEMS-based smart glove that translates hand gestures into both text and speech, making communication more accessible. This smart glove is equipped with flex sensors and MEMS accelerometers that accurately track finger movements and hand positions. A microcontroller processes this data, and a text-to-speech module converts it into spoken words, enabling real-time interaction with those who may not know sign language. Beyond communication, the glove also includes a built-in health monitoring system. It continuously tracks body temperature and heart rate, sending alerts if any abnormalities are detected, ensuring timely medical intervention. Designed to be affordable and easy to use, this wearable technology has the potential to transform the lives of individuals with speech impairments. By making conversations smoother and providing health insights, this smart glove empowers users to engage more freely in daily life. It represents a step toward a more inclusive world where communication barriers are minimized, and everyone has a voice.

25. Advanced Face Mask Detection and Disease Prevention with Image Processing

    M. Moorthi, K. Kishan, Chee Yong Lau, Antony Athithan

    Abstract

    The emergence of global health crises, notably the COVID-19 pandemic, has accentuated the critical importance of preventive measures, one of which is the wearing of face masks. Advanced face mask detection systems, utilizing innovative image processing techniques, have emerged as pivotal tools in disease prevention strategies. These systems not only enhance public health safety but also contribute to the efficient management of infectious diseases in various environments, from crowded urban settings to healthcare facilities. At the core of advanced face mask detection is the application of sophisticated computer vision algorithms, particularly deep learning models. These models are trained on vast datasets comprising images of individuals wearing masks, without masks, and in varying lighting and environmental conditions. By employing convolutional neural networks (CNNs), these systems can accurately identify and classify the presence or absence of face masks on individuals in real-time. This capability is crucial in environments where rapid decision-making is essential, such as airports, schools, and public transport systems.

    Moreover, the integration of image processing techniques not only facilitates mask detection but also enhances overall safety protocols. For instance, systems can be programmed to trigger alerts or record instances of non-compliance, thereby serving as both a deterrent and a monitoring tool. By automating compliance checks, organizations can allocate human resources more efficiently, allowing for a greater focus on other critical health safety measures, thus reinforcing the overall public health response. In addition to the identification of mask usage, advanced image processing systems can be adapted to evaluate other health parameters associated with the spread of infectious diseases. For example, analysing crowd density through video feeds could provide valuable insights into potential risk zones, thereby enabling proactive measures to alleviate viral transmission.

26. Deep Learning Insights into Neurological Brain Disorder Classifications

    M. Moorthi, S. Meghavarshini, J. S. Bhuvaneshwari, C. H. C. Alexander, Antony Athithan

    Abstract

    This paper explores the application of deep learning techniques to classify neurological brain disorders, a vital research domain with significant implications for diagnosis and treatment. A robust framework was designed to effectively classify neurological conditions using medical images, including scans of the brain. The approach integrates multi-modal data fusion, incorporating structural, functional, and diffusion-weighted neuroimaging to comprehensively analyse brain anatomy. The model’s ability to distinguish disorders such as Parkinson’s disease and Alzheimer’s disease was revealed in extensive experiments. Insight into the underlying pathological processes can be offered by interpretability methods, like attention mechanisms. Improved diagnostic tools and personalized therapeutic strategies will benefit patient care and outcomes as a result of these advancements.

27. Backmatter