Mallegowda M
@msrit.edu
MS Ramaiah Institute of Technology Bangalore
EDUCATION
BE,MTech,PhD
RESEARCH, TEACHING, or OTHER INTERESTS
Engineering, Computer Engineering, Computer Science, Artificial Intelligence
Scopus Publications
Scopus Publications
M. Mallegowda, Tejas Hegde, Sini Anna Alex, and Anita Kanavalli
Springer Nature Switzerland
Prasanna Kumar K R, Mallegowda M, Manoj Kumar D P, Swathi H Y, and Ananda Babu J
Informa UK Limited
M Mallegowda, Tresa Maria Josylin, and Anusha M
IEEE
Mallegowda M, M. S. Swaroop, C S Darshan, and Dinesh P
IEEE
Mallegowda M, Nikhlil K Rohidekar, Pavan Reddy T, and Anita Kanavalli
IEEE
M. Mallegowda, A. Parkavi, S. Sanath, Siddharth Satyavolu, and Moulya R. Gowda
Springer Nature Singapore
Mallegowda M, Saanvi Nair, Sahil Khirwal, Manik Animesh, Vasuman Mishra, and Anita Kanavalli
IEEE
Models based on dynamic neural networks such as DeBERTa and Vision Transformers ViT have disrupted the two fields. However, there is one complication – their performance requires so much computation which limits its realistic applications. The aim of this study is to propose advanced optimization techniques that can increase the efficiency of these models without compromising their performance. We adopt approaches such as DistilBERT, DeBERTa and Vision Transformer models. Then we carry out detailed empirical analysis of the consequences of these techniques on the model accuracy, time performance, throughput as well as resources consumption. The findings show enhancement models, where some methods improved the throughput by 29% and reduced the GPU usage by 61% at the same or higher accuracy. This reveals the capacity of dynamic neural networks in resource-constrained settings and enhances the overall vision of the use of such models in practice.
Mallegowda M, Theertha K, Varsha S D, and Anita Kanavalli
IEEE
Large-data set artificial neural network training takes a lot of time. Many ways for reducing effort have been proposed, many of which make use of parallelization techniques. This paper explores the implementation of face detection algorithms utilizing OPENMP to achieve greater efficiency through parallelization. We focus on specific OpenMP parallelization setups that run on a typical multi-threaded CPU. These frameworks are also available for CUDA, however utilizing CUDA is only possible if you have an NVIDIA graphics card, which is clearly not the case for everyone. OpenMP's release of a stable version in late 2015 facilitated parallel processing and broadened its development base.
Subash N, Mallegowda M, S. Rajarajeswari, and Alisha Ahmed
IEEE
In the realm of medical education, the integration of new technologies is pivotal for enhancing learning experiences and improving patient care. Virtual Reality (VR) stands out as a transformative tool, particularly for the instruction of knee surgery. This type of surgery is inherently complex, necessitating a profound understanding of human anatomy and surgical techniques. Traditional teaching methods often fall short in delivering this comprehensive education. However, VR addresses these shortcomings by offering interactive simulations tailored to the specific educational needs of students. It effectively bridges the gap between theoretical knowledge and practical application, facilitating experiential learning and better retention of information. Through VR, students can safely explore human anatomy and practice surgical procedures. Our paper introduces a VR program specifically designed for knee surgery education, utilizing advanced equipment such as the Oculus Quest 2. Our research demonstrates that VR training significantly enhances student engagement, comprehension, and surgical proficiency compared to conventional teaching methods. VR has the potential to revolutionize medical education, making learning more immersive and hands-on, which ultimately leads to better patient outcomes in knee surgery.
N Subash, M Mallegowda, S. Rajarajeswari, and Alisha Ahmed
IEEE
Virtual reality (VR) has emerged as a promising tool in medical education, offering immersive and interactive experiences for trainees to learn complex surgical procedures. This paper explores the development of VR applications specifically tailored for medical education in osteotomy knee surgery. By leveraging VR technology, medical students can engage in realistic simulations of the surgical procedure, enhancing their understanding of anatomy, surgical techniques, and decision-making processes in a controlled and risk-free environment. The VR applications for osteotomy knee surgery providing a comprehensive educational experience. Through a combination of realistic graphics, and interactive scenarios, VR facilitates experiential learning and skill acquisition, ultimately improving the competency. Moreover, the scalability and accessibility of VR platforms offer opportunities for widespread adoption in medical training programs, addressing the growing demand for innovative and effective educational tools in healthcare.
Mohammad Qazim Bhat, D. S. Jayalakshmi, M. Mallegowda, and J. Geetha
Springer Nature Singapore
Mallegowda M, Tanupriya R, Vishnupriya C, and Anita Kanavalli
IEEE
Principal component analysis (PCA) is a crucial technique in data science and machine learning for reducing dimensionality and identifying crucial characteristics. But when dealing with big datasets, it uses Singular Value Decomposition (SVD), which is computationally costly. To address this issue, this study compares the effectiveness of parallel and serial PCA implementations. Serial execution uses a step-by-step data processing method, which is simple but falls short when dealing with large datasets. The task is divided across many processing units in parallel execution, which is preferred in high-performance computing environments, and thus results in significant speed increases for large datasets. While parallel execution shines with huge datasets and time-sensitive activities, serial execution improves with smaller datasets and simple scenarios. Depending on variables like dataset size and available computer resources, one of these techniques may be preferred over the other. In conclusion, this paper examines the advantages of utilizing parallel computing for PCA via SVD, offering a quick way to speed up calculations for high-dimensional datasets. It emphasizes how flexible it is to choose serial or parallel execution depending on the particular dataset's properties and processing needs.
Alisha Ahmed, S. Rajarajeswari, M Mallegowda, and N Subash
IEEE
Advancements in medical education are continuously evolving, with virtual reality (VR) emerging as a groundbreaking technology reshaping the landscape of radiology training. This abstract presents a paradigm shift in radiology education through the creation of immersive VR simulations tailored for medical students and professionals. Our VR simulation leverages state-of-the-art graphics, interactive modules, and realistic patient scenarios to provide a dynamic learning environment. Users can navigate through virtual clinics, interact with lifelike patients, and practice diagnostic procedures in a risk-free yet lifelike setting. The incorporation of haptic feedback and spatial audio enhances realism, fostering deeper engagement and skill acquisition. Furthermore, our VR simulation integrates artificial intelligence algorithms for personalized learning pathways, adapting to individual learning styles and providing real-time feedback. By harnessing the power of VR, this simulation not only enhances diagnostic skills but also promotes critical thinking, teamwork, and decision-making under pressure. This abstract explores the transformative impact of VR simulations in revolutionizing radiology education, paving the way for a new era of experiential learning in medical training.
M. Mallegowda, Sini Anna Alex, Sajal Srivastava, Shashank Singh, and Anita kanavalli
Springer Nature Switzerland
MallegowdaM, Purva Rajodiya, Samruddha S, Sini Anna Alex, and Anita Kanavalli
IEEE
This research proposes an approach to enhance the denoising and upscaling performance of noisy images using Generative Adversarial Networks (GANs), particularly Style GAN architecture. Denoising and upscaling noisy images are crucial in many computer vision applications, and GANs have shown remarkable effectiveness in creating high-quality images. However, training Style GAN requires huge amount of data and is computationally expensive. To address this issue, this study proposes using various filters such as mean, median, and weighted median to pre-process the noisy images before feeding them to Style GAN. The proposed approach achieves superior denoising and upscaling compared with other system in terms of FID and inception score, and further exploration of hyperparameters and variations of the Style GAN architecture can lead to even better results.
M. Mallegowda, Shubeeksh Kumaran, V. Aditya Raj, Skanda S. Kumar, and Ronith H. Gowda
Springer Nature Singapore
Mallegowda M, Vikas Hajjarge, Vinayak Vittal Divate, Krishna Mohan, and Anita Kanavalli
IEEE
In this study, the efficiency of parallel and serial computation techniques for aggregating data from diverse regions is investigated. Parallel computation breaks data into smaller segments and assigns each segment to a different processor or core, as opposed to traditional serial computation, which processes data sequentially. The research focuses on comparing the total amount of time needed by each approach to collect data from diverse regions in order to assess the effectiveness of parallel processing. According to the research, parallel computation significantly reduces the amount of time needed to collect data, and this correlation is directly related to the number of processors or cores used. The potential benefits of multi-core architectures in accelerating the gathering of data from multiple places are highlighted in this study. The outcomes highlight the efficacy of parallel computation methods, illuminating their potential to accelerate multi-regional data collection procedures in a variety of applications.
M. Mallegowda, Anita Kanavalli, Shivalingesh J. Patil, Skanda S. Kumar, Vinayak Vittal Divate, and M. S. Vishnu Patel
Springer Nature Singapore
Mallegowda M, Sanskar R G, VISHVESHWARA N, Safwan G A, Vivek J, and Anita Kanavalli
IEEE
Fruit freshness detection is critical in the food industry for maintaining quality, minimizing waste, and meeting consumer expectations. As the global population grows, preserving fruit freshness is increasingly vital. This project introduces an AI system that identifies fruit types and assesses their freshness using color, texture, and other visual attributes. The system employs Agile methodology and machine learning techniques, such as Convolutional Neural Networks (CNN), separating classification and freshness detection models to handle various fruits under different conditions. Achieving over 90% accuracy and rapid classification, the system works from any device with an internet connection, offering significant benefits to producers and consumers. The use of transfer learning with the VGG16 model and data augmentation techniques highlights the practical application of deep learning. This system enhances operational efficiency in retail, storage, and personal use. Future work will address system limitations by expanding the dataset, improving real-world utility, boosting accuracy with advanced techniques, and incorporating real-time monitoring features.