Kanimozhi Soundararajan

@skanimozhi@vit.ac.in

Assistant Professor (Sr. Gr) and School of Computer Science and Engineering
VIT University, Chennai Campus



              

https://researchid.co/kanimozhis

EDUCATION

Ph.D (ICT)
M.Tech(IT)
B.Tech(IT)

RESEARCH, TEACHING, or OTHER INTERESTS

Multidisciplinary, Computer Engineering, Agricultural and Biological Sciences, Computers in Earth Sciences

21

Scopus Publications

Scopus Publications

  • EffiCAT: A synergistic approach to skin disease classification through multi-dataset fusion and attention mechanisms
    A. Sasithradevi, S. Kanimozhi, Parasa Sasidhar, Pavan Kumar Pulipati, Elavarthi Sruthi, and P. Prakash

    Elsevier BV

  • EVGPT - Fine-Tuned Large Language Model for Electric Vehicle Operational Assistance
    Kanimozhi S., Barath. M, Abel Manoj Mathew, Prithvi G., Vaikash K.G., and Gukhan V.B.S

    IEEE
    The optimization of electric vehicle (EV) utilization and efficiency is becoming increasingly essential in the ever- changing landscape of sustainable transportation. Consulting user manuals, online resources, and seeking expert advice for EV maintenance and operation is a significant obstacle to productivity. Electric Vehicle GPT (EVGPT) is a specialized Large Language Model (LLM) specifically designed to address queries related to EV maintenance, operation, and troubleshooting, filling the gaps that conventional LLMs may miss. This model serves as a comprehensive digital assistant for technicians, operators, and EV enthusiasts, providing timely and precise information while upholding response quality. This study offers an overview of the EVGPT model architecture and presents an operational evaluation based on its implementation in a specific EV model.

  • AI-driven Solutions for Autonomous Maintenance and Fault Detection in Electrical Power Grids
    S. Kanimozhi, Enakshi, Rohan Chhabra, and Shashank Shekhar Rai

    IEEE
    Efficient and reliable power grids are crucial for the functioning of modern society. However, increasing energy demand and renewable sources can lead to power failures. Traditional methods struggle. AI, particularly Random Forest Classifiers, offers a promising approach. These classifiers analyse vast amounts of grid data to identify patterns indicative of faults. This allows for real-time anomaly detection. This research investigates the efficacy of Random Forest models for fault classification, focusing on optimising data preprocessing and collection. Transparency in data collection and model development is crucial for data privacy. While simulated environments pose challenges for early fault detection, real-world data presents even greater complexities. The effectiveness of Random Forest Classifiers in early fault detection is critical for grid reliability, reducing power outages, and enhancing long-term operational efficiency.

  • Real-Time Crop Analysis with Enhanced Image Classification Using Deep Learning and Fuzzy Logic Techniques
    Kanimozhi S, Deepika Roselind J, Neal Shah, and Sai Ramesh L

    Institution of Engineering and Technology (IET)

  • Explainable AI and Deep Learning for Brain Tumour Classification Using Grad-CAM Visualization
    Deepika Roselind Johnson, R. Srivats, Abhiram Sharma, Kalyanasundaram V, and Kanimozhi S

    Institution of Engineering and Technology (IET)

  • Dynamic dashboard creation for sales trends and optimize pricing strategies
    A Sasithradevi, S Kanimozhi, and Srinath Srinivasan

    CRC Press

  • Visualizing food quality and safety: A dynamic dashboard approach with near-infrared imaging and machine learning
    L Brighty Ebenezer, A Sasithradevi, Chanthini Baskar, and S Kanimozhi

    CRC Press

  • Interactive and dynamic stock market dashboard
    Prasan Mittal, S Kanimozhi, and L Sairamesh

    CRC Press

  • Ontology-Based Action Recognition in Sport Video's Using Semantic Verification Model
    S. Kanimozhi, A. Sasithradevi, and L. Sairamesh

    Institute of Electrical and Electronics Engineers (IEEE)
    Sports employ computer vision for a variety of applications, including interactive viewing, object interpretation, sports action analysis, and intelligent rule-based systems for better referee choices. The computer vision problem known as “sports action recognition” looks for interactions between players and related sporting items. Most of the existing solutions in sports action recognition are complex and incomplete as taking both playfield and non-play field objects with various actions for the same human posture into account. Hence, the proposed model focuses only on objects within the playfield and the right functional action that is performed by the human. We identify the play field using a quadrant-based density (QD) approach which locates the crowded region and then morphological operations are applied to spot only play region. The right action performed is identified using the logic-based query to the Sports Action Generation model (SAG) which is generated using the newly constructed sports ontology. The proposed semantic verification model using ontology typically offers more transparency compared to deep learning networks. It relies on explicit rules and logic that can be easily interpreted, enhancing interoperability and accuracy across systems. We use the sports video wild (SVW) and UCF-101 datasets to assess the performance of our proposed model. The proposed sports action recognition system achieves action identification in videos with a good accuracy of 96%.

  • Investigation of Thyroid Nodule Detection Using Ultrasound Images with Deep Learning
    C. Kaushik Viknesh, S. Kanimozhi, and R. Thirumalai Selvi

    IEEE
    the foremost predominant shape of endocrine cancer is thyroid cancer, and its rate has been relentlessly expanding to its worldwide scale. This audit centers on the challenging assignment of distinguishing thyroid knobs utilizing ultrasound images. Ultrasound offers a cost-effective and non-invasive implies of visualizing the thyroid organ and the encompassing tissues. Truly, consideration has transcendently coordinated to experimentally approve imaging highlights utilized in ultrasonography for thyroid knob discovery. The essential goal of computerized knob location is to achieve a level of precision that rivals that of fine needle goal biopsy. In light of this objective, we have presented an inventive approach for knob acknowledgment, tackling the control of convolutional neural systems. Convolutional neural systems (CNNs) have a particular advantage in naturally learning high-level and various leveled reflections from visual information through end-to-end preparing. The adequacy of knob distinguishing proof is too surveyed from three diverse points: multiscale prediction engineering, post-processing strategies, and plans of the loss work. Clinical information assesses the execution of this approach, with comparisons made to the ground truth labels given by restorative experts. Moreover, when a thyroid knob is recognized, it gets to be conceivable to determine whether the condition has spread to adjacent zones, potentially driving to progressed persistent results. In rundown, the executions of our proposed strategy holds significant potential to upgrade the viability and accuracy of thyroid knob discovery and characterization.

  • Investigation of IOT Enabled Maternal Health Monitoring System for Antenatal Care
    S. Kanimozhi and C. Kaushik Viknesh

    IEEE
    In developing nations, a significant portion of the population resides in rural regions where healthcare systems lack integration for information sharing. Mainly, expectant mothers struggle to undertake regular prenatal checkups, leading to elevated infant and maternal mortality rates in both rural and urban settings. This scenario presents substantial healthcare challenges for women. To address this, an accelerometer sensor is devised to gauge the strength and frequency of fetal movements. This data is then transmitted to a RASPBERRY PI Pico controller. The motion of the fetus and vital metrics like blood pressure, heart rate, fetal kicks count, and maternal temperature are monitored through an array of diverse sensor technologies. The gathered data is communicated via IOT and showcased on mobile devices. This setup, with its heightened sensitivity and lightweight design capable of detecting even subtle motions, is well-suited for domestic monitoring. Contrary to the prevailing trend of employing costly and extended-use ultrasound scans, which exhibit unclear fetal impact limitations, this approach discourages their continuous utilization. Moreover, due to its high cost, we have put forth an alternative system incorporating a range of sensors including a heartbeat sensor, temperature sensor, blood pressure sensor, and an accelerometer sensor to monitor fetal movements. These sensors enable data collection, which is then transmitted to a mobile application via Raspberry pi Pico and IOT technology. Should any irregular readings arise, GSM module is employed to compute normal and abnormal rates.

  • Breast Cancer Histopathological Image Classification Using CNN and VGG-19
    S. Kanimozhi and S. Priyadarsini

    IEEE
    Applications of deep learning in medicine, like iden- tifying the kind of malignant cells, are common. Breast cancer is a most common type of cancer in women and it a main cause of death for women. There are three categories for the malignant cells: Normal, Mild, and Severe. Early diagnosis of the malignant cells can prevent these deaths. Numerous techniques, including MRIs, mammograms, ultrasounds, and biopsies, are used to identify cancerous cells. Hematoxylin and eosin-stained breast cancer histology photos are difficult to diagnose, labor-intensive, and frequently cause pathologists to disagree. Recent advances in deep learning have made histological image processing possible with convolutional neural networks (CNNs). Histology images of breast cancer are categorized into sub-classes based on general tissue structure and morphology, as well as the density, variabil- ity, and organize of the cells. These subclasses include benign, malignant, and normal. Using this information, extract features at the cell and tissue levels, respectively from histopathological images, in smaller and larger size patches. The dataset repository is where the input image was obtained. The image has to be pre- processed. The feature extraction must then be put into practice. The pre- processed image must then be segmented. The image must be divided. We are able to apply many neural network models, including VGG-19 and Convolutional Neural Network (CNN). The findings of the experiment indicate that the accuracy. The primary goal of our method is to identify or anticipate breast cancer based on the input image.

  • Interactive Learning Environment for Networking Topics based on Augmented Reality
    S. Kanimozhi and L. Sairamesh

    Institution of Engineering and Technology

  • Automatic Lung Disease Detection and Diagnosis Using Optimized Fuzzy Filter and Deep Learning Method
    Senthil Pandi S, Kanimozhi S, Rahul Chiranjeevi V, and Ishwarya M

    IEEE
    Corona Virus Disease (COVID-19) is a life-threatening disease that was found in December of 2019 in Wuhan, China. A quick prediction can isolate infected people and prevent the disease from spreading to others. For some people, the Reverse Transcription Polymerase Chain Reaction (RT-PCR) test yields the negative result, but they may have an infection, so CT image prediction is the best solution for patients with symptoms. The primary goal of this research is to detect lungs infection quickly and automatically, as well as to improve COVID-19 prediction accuracy by improving image quality through image restoration and enhancement. The proposed research is divided into two modules: the first explains how to improve the quality of input lungs Computerized Tomography (CT) images using an optimized fuzzy filter, and the second explains how to predict and diagnose COVID-19 using DenseNet-121. The investigational results show that the proposed approach can accomplish 0.96 AUC score, 0.88 precision, 0.92 recall, 0.90 accuracy and 0.259 model loss. The proposed method outperforms previous works in CT images. Experimental results indicate that introducing an optimized fuzzy filter, which is used for image restoration and enhancement, improves prediction accuracy.

  • Agricultural Crop Recommendation Based on Productivity and Season
    Umamaheswari S, Aartisha S, Kanimozhi S, Birnica Y J, and Suhashini R

    IEEE
    Machine learning plays a major role in every sector. Machine learning makes the work easier and more accurate. A significant source of employment is farming in Tamil Nadu. Agriculture production is suffering because of the changes in the environment. The factors include humidity, rainfall, sunlight, soil type, and temperature do have a direct impact on agriculture. Agriculture need proper knowledge for crop harvesting is required to flourish. The information used to get awareness into the Agri- facts are created by agricultural metrics and factors. By using several machine learning techniques, we can predict the crops and create a model from the given data. This might enable aspiring farmers to practice better agriculture. By the help of data mining, the crop production is increased thus by recommending the farmers about the crops. Crops are suggested for use in such a strategy depending on their quantity and meteorological considerations. Data analytics opens the door for the development of valuable agricultural database extraction. The analysis of the crop dataset resulted in the crop recommendation based on their productivity and growing season.

  • Sports highlight recognition and event detection using rule inference system
    Kanimozhi Soundararajan and Mala T

    SAGE Publications
    Computer vision in sport is a very interesting application. People spend a lot of time watching sports videos because this is one of the best field of entertainment. Sports video broadcasts generally take a lot of time, ranging from two to four hours. However, the interesting part happens for just a few minutes. Detecting the highlighted event in a sport will be useful for people who like to watch only the prominent events section instead of watching the whole video broadcast. Event detection will give precise details about the action that occurred for a particular time, but the detection of highlighted events is more complex. This is due to the fact that a sports video contains collections of events. Among them, segregation of the required event is a time-consuming process but it requires more knowledge about the sport as well as processing time. Hence, a novel work is proposed focused on identifying the location of the functional object using agglomerative clustering and annotating the event highlights automatically by means of the rule inference mechanism. The SHRED (Sports Highlight Recognition and Event Detection) system achieves an overall accuracy of about 97.38% relative to other state-of-art methods in event class annotation.

  • Abnormality Detection in Human Action Using Thermal Videos
    Kanimozhi Soundararajan, Anbarasi Soundararajan, and Sai Ramesh L

    IOS Press
    Anomaly detection is a challenging task in the surveillance system due to the factors like extracting appropriate features, inappropriate differentiation among the normal vs abnormal behaviours, the sparse occurrence of abnormal activities and environmental variations. In the dark environment, detection of human actions is still difficult as more features for recognizing the key point are not visible. Hence the proposed work is focused on overcoming the environmental variations task that too in a less bright environment by using thermal videos. Variations in the actions can be easily identified as it works on the property of infrared radiations. For recognizing actions, the skeleton-based approach is used as it helps with the joint-wise segregation of human parts, resulting in more accuracy. The motion pattern of humans in the thermal video is tracked to classify the level of abnormality.

  • Key Object Classification for Action Recognition in Tennis Using Cognitive Mask RCNN
    S. Kanimozhi, T. Mala, A. Kaviya, M. Pavithra, and P. Vishali

    Springer Singapore

  • Captioning of Image Conceptually Using BI-LSTM Technique
    Thaseena Sulthana, Kanimozhi Soundararajan, T. Mala, K. Narmatha, and G. Meena

    Springer International Publishing

  • Distinct actions classification using human action tracker technique in sports videos
    Kanimozhi S, Anbarasi S, and Mythili M

    IOS Press
    Recognizing human action in sports is difficult task as various sequences of activities involved in every scene. Identifying each action individually without overlapping of movements is a tedious process due to continuous change of frames within short duration. So proper tracking of human movements for each action is important. Hence new structure-based human action recognition and tracker technique (HART) is proposed. It uses joint trajectory images and visual feature to design each human action. At first, a structural based method employed to extract human skeleton data points from RGB (Red Green Blue) videos. Next, a Multitude Object Tracker (MOT) is proposed which uses the trajectory of human skeleton joints in an image space for identification of actions. Then, Histogram of Oriented Gradients (HOG) combined with Support Vector Machine (SVM) is applied to extract physical body shape and action information. Finally, the action label and interconnected keypoints in humans is jointly detected as end result. The proposed HART technique effectively performed well with the accuracy of about 82% over the other activity recognition methods.

  • Multiple real-time object identification using single shot multi-box detection
    S Kanimozhi, G Gayathri, and T Mala

    IEEE
    Real time object detection is one of the challenging task as it need faster computation power in identifying the object at that time. However the data generated by any real time system are unlabelled data which often need large set of labeled data for effective training purpose. This paper proposed a faster detection method for real time object detection based on convolution neural network model called as Single Shot Multi-Box Detection(SSD).This work eliminates the feature resampling stage and combined all calculated results as a single component. Still there is a need of a light weight network model for the places which lacks in computational power like mobile devices( eg: laptop, mobile phones, etc). Thus a light weight network model which use depth-wise separable convolution called MobileNet is used in this proposed work. Experimental result reveal that use of MobileNet along with SSD model increase the accuracy level in identifying the real time household objects.

RECENT SCHOLAR PUBLICATIONS

    Publications

    Kanimozhi, S., Gayathri, G., & Mala, T. (2019, February). Multiple Real-time object identification using Single shot Multi-Box detection. In 2019 International Conference on Computational Intelligence in Data Science (ICCIDS) (pp. 1-5). IEEE.
    Kanimozhi, S., Mala, T., Kaviya, A., Pavithra, M., & Vishali, P. (2022). Key Object Classification for Action Recognition in Tennis Using Cognitive Mask RCNN. In Proceedings of International Conference on Data Science and Applications (pp. 121-128). Springer, Singapore.
    Soundararajan, K. (2022). Sports highlight recognition and event detection using rule inference system. Concurrent Engineering, 1063293X221088353.
    Sulthana, T., Soundararajan, K., Mala, T., Narmatha, K., & Meena, G. (2021, March). Captioning of Image Conceptually Using BI-LSTM Technique. In International Conference on Computational Intelligence in Data Science (pp. 71-77). Springer, Cham.