Dr.T.Kamaleshwar

@veltech.edu.in

Assistant Professor, Computer Science and Engineering
Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology

Dr.T.Kamaleshwar


            Download Compact PDF  
https://researchid.co/kamalesh4u2

RESEARCH INTERESTS

Networking, Digital Image processsing
24

Scopus Publications

Scopus Publications

  • AGRINEXUS: The Hub for Agriculture and Financial Services
    Adusumalli Rithika, T. Kamaleshwar
    Harnessing AI to Reshape the Future of Agriculture, 2026
  • Lightweight Object Detector for Pest Detection on Crops Under Field Conditions
    Amandeep Kaur, Sathish K, Dilli Ganesh V, T.Kamaleshwar, Harishchander Anandaram, Niranjanamurthy M
    Proceedings of 4th International Conference on Electronics and Renewable Systems Icears 2026, 2026
  • Rwetraction note: Deep neural network with reduced feature for classification of breast cancer mammogram (Soft Computing, (2022), 26, 24, (14021-14028), 10.1007/s00500-022-07533-7)
    N. N. Krishna Veni, V. Preetha, K. Meena, T. Kamaleshwar, A. V. R. Mayuri, Shareefunnisa Syed
    Soft Computing, 2026
  • AI-Powered Web Application for Smart Vehicle Sharing
    T. Kamaleshwar, Thota Sai Lakshmi Lalasa, Darsi Navya Sri
    Proceedings of 6th International Conference on Iot Based Control Networks and Intelligent Systems Icicnis 2025, 2025
    Ridezipp is an exclusive, online intercity ride-sharing service that is intended to resolve the safety, trust, and seating integrity issues of the long-distance transportation. Specifically, the system targets the trips that have a span of over 100 km and incorporates gender-oriented approach towards the distribution of seats in which women are put at the forefront as the system allows individuals to share a transport to save money. In contrast to traditional ride-sharing applications, Ridezipp applies rigid rule-based logic to the visibility of ride, cut-off times of bookings, payment authentication and seat assignment, which guarantees predictable and understandable system behavior. The platform is deployed with a lean and scholar-friendly technology stack based on HTML, CSS and Vanilla JavaScript as the frontend, Python as the backend processing engine and SQLite to store and manipulate data. Such important functional modules are secure authentication, including OTP support, driver authentication, such as license upload, controlled ride creation/search, manual QR-based payment verification, gender-aware seat assignment at the booking close time, credit of reward points, and ride history tracking. The system incorporates support AI elements including OCR-based extraction of transaction IDs when shown on payment screens and analysis of screenshot similarities to identify fraud to increase usability and security. Ridezipp offers a scalable, convenient, and safety-first design approach to intercity ride sharing with the help of rule-based automation and selective AI assistance to ensure reliability and provide users with a comfortable experience. The proposed system can help in understanding how smart constraints and specific AI implementation can enhance trust, performance, and customer experience in shared mobility systems to a large magnitude.
  • Anime Recommendation System using Machine Learning and React
    Bhaskaruni Harini, Gangula Gowtham Reddy, T. Kamaleshwar
    Proceedings of 6th International Conference on Iot Based Control Networks and Intelligent Systems Icicnis 2025, 2025
    In today’s digital media ecosystem, it can be difficult for audiences to determine which anime shows would work well for them with the rise of anime series and availability for viewing. In this paper, we recommend a hybrid system consisting of machine learning (ML), data processing, and a React-based web interface, to provide smart and personalized predictions for anime recommendations. The proposed system collects user data from previous viewing histories , and previously assigned ratings and genre preferences and then applies ML models to identify trends in user behavior and ultimately create a model to estimate the user’s model. The backend uses content-based filtering and collaborative filtering techniques to generate scores of similarity between users and anime, and ultimately generate recommendations for each user. The front end, built in React, provides an interactive and responsive interface for users to explore, rate, and receive recommendations in real time. The model also incorporates continuous feedback to make accurate predictions in real time. This research contributes to the field by (1) applying ML algorithms to personalize anime recommendations from user behavior and content similarity, (2) offering a responsive React interface for interaction in real-time, and (3) showcasing an adaptive environment for recommendations that improve user draw in digital entertainment. Our model demonstrates the opportunity for efficiency, scalability, and design in machine learning in conjunction with upgraded web construction technology to generate user-centered recommendation systems.
  • Anomaly Detection in Crowd Behavior for Public Safety Assurance
    Kamaleshwar T, Dunesh Kumar Madasu, Vishnu Pranavi Damacharla
    Proceedings of 6th International Conference on Iot Based Control Networks and Intelligent Systems Icicnis 2025, 2025
    Public ceremonies, such as events, classes, places of worship, or exhibitions, often include a high concentration of people in limited indoor places. In such an environment, health issues, imbalances or conflicts increases the risk of accidental human fall. Traditional monitoring and manual monitoring systems fail to immediately detect or provide an alert for such important events, leading to delayed reactions that can spoil the results of the injury or even cause congestion nervousness. Addressing this challenge, our project proposes a real-time software solution that takes advantage of AI and Computer Vision to detect human falls and immediately informs security personnel or carers for timely intervention. The system uses a live video feed from existing CCTV or IP cameras and applies advanced deep learning models such as YOLOv 8 for human identity jointly with pose assessment framework to track body movement. When the sudden loss of posture or horizontal decline is detected, the system sends an automated alert through integrated services such as firebase cloud messaging or Twilio SMS API. Alerts can be configured to access mobile devices, web dashboards or alerting systems in real time, which ensures rapid response from on-ground personnel. Major features of the proposed system include monitoring non-grace (no need for wearable equipment), compatibility with existing infrastructure, scalability and congestion settings at many places include high accuracy (tests with 40–50 individuals per room). The modular architecture of the system includes video stream handling, fall detection engine, alert module and monitoring interfaces for logging and visualization. This solution increases safety in public places by reducing dependence on manual observation and enabling early emergency response. The project shows how artificial intelligence can be applied for social good by improving mob safety and real-time phenomenon. Future work can focus on external landscapes, obstacles handling and multi-camera integration to further improve reliability and coverage.
  • Smart Solar Harvesting and Power Management in IoT Nodes Through Deep Learning Models
    Angotu Nageswara Rao, Ravi Shankar Garapati, R.T. Suganya, A. Kaliappan, T. Kamaleshwar, Pranav. N
    2nd International Conference on Intelligent Algorithms for Computational Intelligence Systems Iacis 2025, 2025
    Sensors attached to IoT devices have substantially boosted the amount of data collected about the surroundings. Better ecological understanding and resource management are made possible by the decreasing cost of monitoring systems brought about by quicker gearbox speeds and lower power utilization. Power management and efficient solar collection are essential for remote IoT nodes. To maximize energy efficiency in Internet of Things nodes, they offer a GRU-TCN model, which combines a Gated Recurrent Unit with a Temporal Convolutional Network. Time-series solar irradiance data is used by GRU to extract temporal patterns, and meteorological parameters across the target and neighbouring regions are used by TCN to obtain spatial correlations. Following data cleansing, normalisation, and time-series transformation, features are retrieved. Univariate and multivariate GRU-TCN models were compared to baseline GRU and TCN models using two assessment measures. The proposed GRU-TCN model is quite accurate in its predictions and has a precision of 97.17 percent. Solar harvesting, energy efficiency, and the longevity of IoT-based environmental monitoring systems can all are enhanced using advanced deep learning architectures, according to these studies.
  • Apply the Multi-Agent Reinforcement Learning (RL) Techniques for QoS-Aware Resource Allocation in 5G Networks
    T Sanjeeva Rao, T. Kamaleshwar, Thaer Ahmad Abu-Saleem, Alok Jain, V. Manohari, Uma Maheswari S
    2025 IEEE Pune Section International Conference Punecon 2025, 2025
    Recently, the advent of 5G networks has introduced unprecedented high data rates, ultra-low latency, and massive device connectivity. Due to the mobility of the traffic, the heterogeneous requirements of users and services pose a serious challenge in terms of resource allocation. Classical optimization approaches have limitations in dynamically adapting to these complexities in real time and, at the same time, guaranteeing Quality of Service (QoS) of different applications. This work presents an approach, based on Multi-Agent Reinforcement Learning (MARL), to learn for intelligent QoS-aware resource allocation in 5G networks. In the considered scenario, several agents perform decentralized or cooperative resource allocation of the physical layer (i.e., bandwidth, spectrum, transmission power). Each agent, organized in a decentralized way, is in charge of resource management in its own cell (self-HETNet) or network slice (SCHC), learning how to allocate optimally in the environment iteratively. The agents use reinforcement learning algorithms including Deep Q-Networks (DQN), Policy Gradient techniques and Actor-Critic designs to change their allocation decisions instantaneously, according to network states and QoS performance. The framework of MARL allows partial observations be centralized and also permits the learning of coordinating policies to maximize the global network performance while fairness among users is preserved. QoS constraints (latency, throughput, packet loss rate) are integrated into the reward functions, and the learning process is conditioned to enable the learning of application-specific needs for services like eMBB, URLLC, and mMTC. The simulation results show that the MARL-based resource allocation surpasses the traditional centralized optimization and single-agent RL. The proposed scheme can also adapt quickly to changes in network conditions, and reduce as much as 35% the latency, and increase the spectral efficiency by as high as 25%, while providing reliable QoS delivery to various categories of users simultaneously. Another advantage of MARL is its distributed nature, which leads to low signaling overhead and potential for scalability- features that can be utilized for dense 5G deployments and future 6G networks. The research effort makes a step toward closing the chasm between AI-centric decision-making and the down to earth 5G network management. With the MARL, the 5G infrastructure can realize autonomous, intelligent and scalable resource allocation policies that simultaneously meet the operator's goals as well as the QoS expectation of the end-users, which is ushering in the age of highly flexible and self-organized wireless communication systems.
  • Novel Ml-Based Traffic Flow Optimization: Real-Time Vehicle Monitoring and Signal Management with Internet of Things-Based Metropolitan Infrastructure
    P. Arunkumar, Ch Srivardhan Kumar, T. Pandiselvi, D. Shanmugam, V. Samuthira Pandi, T. Kamaleshwar, Sagunthala
    Proceedings of the 2025 3rd International Conference on Advances in Computation Communication and Information Technology Icaiccit 2025, 2025
    Urban gridlock continues to be a significant challenge facing cities because it leads to loss of man-hours, pollution, and a lower quality of life of its citizens. Conventional traffic signal control systems use fixed-time coordinated sets or minimal sensor input and therefore do not adequately address real-time dynamic changes in traffic. Hence, this study presents a new ML-based framework for traffic flow optimization, which combines real-time vehicle monitoring with adaptive signal control of an IoT-based metropolitan infrastructure. The system utilizes an IoTinfrastructure with sensors, cameras, and connected vehicles, receives data on the (almost) continuous state of traffic and applies ML algorithms to predict the pattern of congestion and automatically adjust traffic signals. The proposed framework leverages the state-of-the-art ML models such as RL for decision making and DL for pattern recognition to accommodate large-scale and diverse traffic data. IoT network infrastructure allows road sensors, traffic lights and control centers to interact seamlessly and function as a responsive traffic management system. The surveillance system uses computer vision for vehicle detection and classification and performs on edge computing nodes with the object of data processing on the edge side to minimize bandwidth requirements and reduce latency. The system was tested in a simulated city traffic scenario with real scanning measurements and traffic data. Simulation results show that average vehicle waiting time can be decreased by as much as 35%, while intersection throughput can be increased by 25% as compared with traditional fixed-timing method. The model also suits well for quick changes in traffic, i.e. for accidents, roadwork's, weather disturbances, and it performs much better than common adaptive traffic systems. In addition to mitigating congestion, the model facilitates environmental gains by cutting idle time and fuel usage, and therefore lowering <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{CO}_{2}$</tex> emission. It also paves the way for interconnection with automated traffic systems and smart city infrastructures, where vehicle and signal coordination will grow increasingly important. This research tackles the challenges of data fusion, real-time decision-making and IoT scalability, and introduces a scalable and flexible solution for traffic optimization within a metropolitan area.
  • Intrusion detection based on phishing detection with machine learning
    R. Jayaraj, A. Pushpalatha, K. Sangeetha, T. Kamaleshwar, S. Udhaya Shree, Deepa Damodaran
    Measurement Sensors, 2024
    Machine learning technique which uses artificial neural networks to learn representations. Phishing is a form of fraud in which the attacker tries to learn credential information from the websites. Web phishing is to steal sensitive information such as usernames, passwords and credit card details by way of impersonating a authorized entity. The Hybrid Ensemble Feature Selection is a new feature selection method for machine learning-based phishing detection systems (HEFS). The first step of HEFS involves using a novel Cumulative Distribution Function gradient (CDF-g) algorithm to generate primary feature subsets, which are then fed into a data perturbation ensemble to generate secondary feature subsets. We present the results of our approach and compare them to a few previous studies, with the paper focusing primarily on phishing urls for detecting the unauthorised one by using phishing detection method.
  • Lightweight Cryptography Model for Overhead and Delay Reduction in the Network
    Rajesh Yamparala, T. Kamaleshwar
    Lecture Notes of the Institute for Computer Sciences Social Informatics and Telecommunications Engineering Lnicst, 2024
  • Breast Cancer Classification Using Synthesized Deep Learning Model with Metaheuristic Optimization Algorithm
    Selvakumar Thirumalaisamy, Kamaleshwar Thangavilou, Hariharan Rajadurai, Oumaima Saidani, Nazik Alturki, Sandeep kumar Mathivanan, Prabhu Jayagopal, Saikat Gochhait
    Diagnostics, 2023
  • A Robust Blind Watermarking Technique for Improved Medical Image Security
    A. Shankar, S. Yazhinian, T. Kamaleshwar
    2023 International Conference on System Computation Automation and Networking Icscan 2023, 2023
  • Auction based Resource Allocation in Cloud Computing using Blockchain Techniques
    14th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2023, 2023
  • Deep neural network with reduced feature for classification of breast cancer mammogram
    N. N. Krishna Veni, V. Preetha, K. Meena, T. Kamaleshwar, A. V. R. Mayuri, Shareefunnisa Syed
    Soft Computing, 2022
  • Deep Learning Application of Image Recognition Based on Self-driving Vehicle
    Stuti Bhujade, T. Kamaleshwar, Sushma Jaiswal, D. Vijendra Babu
    Communications in Computer and Information Science, 2022
  • Decision Trees to Detect Malware in a Cloud Computing Environment
    Vijayaraj, Sumathi M, Rajkamal M, Uganya Vijayaraj, DT. Kamaleshwar, D. Rajalakshmi
    Proceedings of the 2022 International Conference on Electronic Systems and Intelligent Computing Icesic 2022, 2022
  • RETRACTED: Self-Adaptive Framework for Rectification and Detection of Black Hole and Wormhole Attacks in 6LoWPAN
    T. Kamaleshwar, R. Lakshminarayanan, Yuvaraja Teekaraman, Ramya Kuppusamy, Arun Radhakrishnan
    Wireless Communications and Mobile Computing, 2021
  • Secure intrusion detection system in mobile Ad Hoc networks using effective hybrid Cryptographic Technique
    T Kamaleshwar, K Venkatachlapathy
    Journal of Computational and Theoretical Nanoscience, 2019
  • Improving the node efficiency using symmetric key cryptographic hash function technique
    Journal of Advanced Research in Dynamical and Control Systems, 2019
  • Retraction Notice: Improving the Node Efficiency Using Cryptographic Technique (Proceedings of the 2nd International Conference on Trends in Electronics and Informatics, ICOEI 2018 DOI: 10.1109/ICOEI.2018.8553742)
    T. Kamaleshwar, K. Venkatachalapathy
    Proceedings of the 2nd International Conference on Trends in Electronics and Informatics Icoei 2018, 2018
  • Retracted: Improving the Node Efficiency Using Cryptographic Technique (2018 2nd International Conference on Trends in Electronics and Informatics (ICOEI) DOI: 10.1109/ICOEI.2018.8553742)
    T. Kamaleshwar, K. Venkatachalapathy
    Proceedings of the 2nd International Conference on Trends in Electronics and Informatics Icoei 2018, 2018
  • Improving the Node Efficiency Using Cryptographic Technique
    T. Kamaleshwar, K. Venkatachalapathy
    Proceedings of the 2nd International Conference on Trends in Electronics and Informatics Icoei 2018, 2018
  • An effective framework with n-client transfer dataset for weather prediction using data mining techniqueS
    K. Venkatachalapathy, T. Kamaleshwar, D. Sundaranarayana, V. Ohm Prakash
    ACM International Conference Proceeding Series, 2016