Deepa Karuppaiah

Deepa Karuppaiah

Auto-generated profile, based on publicly available information

Deepa Karuppaiah

Vellore Institute of Technology

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 February 2025 in IEEE Access
Verified
Journal Article
Open access
PPPal Patel
Pal Patel
SPShreyansh Pampaniya
Shreyansh Pampaniya
AGAnanya Ghosh
Ananya Ghosh
RRRitu Raj
Ritu Raj
DKDeepa Karuppaih
Deepa Karuppaih
SKSaravanakumar Kandasamy
Saravanakumar Kandasamy
Assistive technologies powered by machine learning are transforming the way sensory impairments are addressed, offering innovative solutions for individuals with hearing and visual disabilities. This paper provides a comprehensive review of machine learning algorithms designed to enhance accessibility for these groups. For hearing impairments, the analysis focuses on advanced models such as Support Vector Machines (SVM), Random Forests (RF), and Multi-Layer Perceptrons (MLP), examining their effectiveness in auditory assistive applications. In the context of visual impairments, state-of-the-art object detection frameworks like You Only Look Once (YOLO), Single Shot MultiBox Detector (SSD), and RetinaNet are evaluated for their capability to enable real-time object recognition and navigation aids. The study also reviews the Generative Artificial Intelligence based applications for visual and hearing impaired use cases. The study addresses the unique challenges and requirements associated with each type of sensory impairment, with particular emphasis on the customization and fine-tuning of machine learning models for personalized, effective solutions. Additionally, it highlights the transformative potential of deep learning models in advancing assistive technologies, ultimately aiming to enhance the quality of life for individuals with sensory disabilities. By advocating for the development and integration of such technologies, this paper underscores the importance of inclusivity and empowerment in creating a more equitable society.

26 September 2024 in IEEE Access
Verified
Journal Article
Open access
AGAnanya Ghosh
Ananya Ghosh
KDK. Deepa
K. Deepa
QueryMintAI, a versatile multimodal Language Learning Model (LLM) designed to address the complex challenges associated with processing various types of user inputs and generating corresponding outputs across different modalities. The proliferation of diverse data formats, including text, images, videos, documents, URLs, and audio recordings, necessitates an intelligent system capable of understanding and responding to user queries effectively. Existing models often exhibit limitations in handling multimodal inputs and generating coherent outputs across different modalities. The proposed QueryMintAI framework leverages state-of-the-art language models such as GPT-3.5 Turbo, DALL-E-2, TTS-1 and Whisper v2 among others, to enable seamless interaction with users across multiple modalities. By integrating advanced natural language processing (NLP) techniques with domain-specific models, QueryMintAI offers a comprehensive solution for text-to-text, text-to-image, text-to-video, and text-to-audio conversions. Additionally, the system supports document processing, URL analysis, image description, video summarization, audio transcription, and database querying, catering to diverse user needs and preferences. The proposed model addresses several limitations observed in existing approaches, including restricted modality support, lack of adaptability to various data formats, and limited response generation capabilities. QueryMintAI overcomes these challenges by employing a combination of advanced NLP algorithms, deep learning architectures, and multimodal fusion techniques.

9 December 2024 in IEEE Access
Verified
Journal Article
Open access
SCSanika Chavan
Sanika Chavan
JGJahnavi Gundakaram
Jahnavi Gundakaram
SVSai Dyuti Vaishnavi
Sai Dyuti Vaishnavi
SPSrishti Prasad
Srishti Prasad
KDK. Deepa
K. Deepa
Cryptocurrency markets exhibit high volatility, necessitating accurate forecasting methods for effective decision-making. This paper presents an innovative approach that integrates web scraping from cryptocurrency websites with various deep-learning networks to predict cryptocurrency values for the following day. Our web scraping technique integrated with concept like multi-threading focuses exclusively on cryptocurrency websites, extracting essential data such as live price records making use of crucial computer technology concepts like multi-threading. Combined with a suite of deep learning models including LSTM, GRU, and XgBoost, this data facilitates the modelling of temporal dependencies crucial for understanding cryptocurrency price dynamics. Through empirical evaluation, we determine the model that outperforms others and integrate it into a dashboard for real-time prediction. By leveraging real-time insights from web scraping, our model aims to enhance prediction accuracy. This research contributes to the advancement of predictive analytics in cryptocurrency trading, providing actionable insights for investors and analysts amidst fluctuating market conditions.

21 February 2024 in E3S Web of Conferences
Verified
Conference Paper
Open access
PDF file
SKShashwat Kumar
Shashwat Kumar
AKArchisa Kumar
Archisa Kumar
DGDisha Goyal
Disha Goyal
ACAnannya Chuli
Anannya Chuli
RMRiddhi Maniktalia
Riddhi Maniktalia
KDK. Deepa
K. Deepa
According to estimates, every year 10% of global production, goes waste due to pests and crop pathogens. For instance, India is a leading producer of many crops, including wheat, rice, lentils, sugarcane, and cotton. But a majority of the farmers are unable to detect whether a crop is infected or not simply by looking at it. As crop pathogens develop greater resistance to fungicides and pesticides, there is an urgent need to find new antifungal compounds to effectively combat them, which over time are rendered useless as the pathogens again develop resistance to these compounds. Thus, the food security of any country is always at risk due to the vulnerability of the current agricultural systems to climate, pests, pathogens, and associated diseases. To solve this problem, we have developed Cropable, The Crop Protection App. In the proposed work, we have used Deep Convolution Neural Networks( CNN) models to detect the disease and further created a web app using flask. Cropable is an Artificially Intelligent Web Application that can help to identify whether the crop is infected or not. We also provide farmers with a treatment for the detected disease, which not only helps them in identifying a disease but also assists them in solving it.

28 April 2016 in Indian Journal of Science and Technology
Verified
Journal Article
Open access
PDF file
SKSaravanakumar Kandasamy
Saravanakumar Kandasamy
KDK. Deepa
K. Deepa
Internet users have grown in recent years and they demand answers for many through online. Searching and retrieving documents is one of the most frequent thing most of the people do today. For retrieving medical related documents, we have to be extra careful and precise in the process of retrieval. Even though separate search engines are there to retrieve medical documents with care, the users are not well-known with MeSH terms (Medical Subject Heading). MeSH terms are terms used by medical professionals to retrieve documents. So the query has to be framed in such a way that only correct documents should be produced to the user. In this work, we proposed a method that deals with enriching the user query with the use of UMLS. The enriched user query after expansion is used to get accurate documents with less amount of time.