BITS Faculty Publications

Permanent URI for this communityhttp://localhost:4000/handle/123456789/1867

Browse

Search Results

Now showing 1 - 10 of 21

Medical image segmentation using advanced UNETt: VMSE-Unet and VM-Unet CBAM+
(2025-07) Chalapathi, G.S.S.
In this paper, we present the VMSE U-Net and VM-Unet CBAM+ model, two cutting-edge deep learning architectures designed to enhance medical image segmentation. Our approach integrates Squeeze-and-Excitation (SE) and Convolutional Block Attention Module (CBAM) techniques into the traditional VM U-Net framework, significantly improving segmentation accuracy, feature localization, and computational efficiency. Both models show superior performance compared to the baseline VM-Unet across multiple datasets. Notably, VMSEUnet achieves the highest accuracy, IoU, precision, and recall while maintaining low loss values. It also exhibits exceptional computational efficiency with faster inference times and lower memory usage on both GPU and CPU. Overall, the study suggests that the enhanced architecture VMSE-Unet is a valuable tool for medical image analysis. These findings highlight its potential for real-world clinical applications, emphasizing the importance of further research to optimize accuracy, robustness, and computational efficiency.
Transformers for vision: a survey on innovative methods for computer vision
(IEEE, 2025-05) Kumar, Dhruv; Chalapathi, G.S.S.
Transformers have emerged as a groundbreaking architecture in the field of computer vision, offering a compelling alternative to traditional convolutional neural networks (CNNs) by enabling the modeling of long-range dependencies and global context through self-attention mechanisms. Originally developed for natural language processing, transformers have now been successfully adapted for a wide range of vision tasks, leading to significant improvements in performance and generalization. This survey provides a comprehensive overview of the fundamental principles of transformer architectures, highlighting the core mechanisms such as self-attention, multi-head attention, and positional encoding that distinguish them from CNNs. We delve into the theoretical adaptations required to apply transformers to visual data, including image tokenization and the integration of positional embeddings. A detailed analysis of key transformer-based vision architectures such as ViT, DeiT, Swin Transformer, PVT, Twins, and CrossViT are presented, alongside their practical applications in image classification, object detection, video understanding, medical imaging, and cross-modal tasks. The paper further compares the performance of vision transformers with CNNs, examining their respective strengths, limitations, and the emergence of hybrid models. Finally, current challenges in deploying ViTs, such as computational cost, data efficiency, and interpretability, and explore recent advancements and future research directions including efficient architectures, self-supervised learning, and multimodal integration are discussed.
Enhancing data privacy: a comprehensive survey of privacy-enabling technologies
(IEEE, 2025-02) Chalapathi, G.S.S.
Privacy is a fundamental human right, especially crucial in our modern digital age. With the rapid advancement of technology, ensuring individuals’ privacy has become increasingly complex. Our survey paper aims to shed light on various privacy engineering technologies that play a crucial role in protecting personal data. We delve into four key areas: data anonymization, data encryption, synthetic data generation, and differential privacy. These technologies serve as essential tools in safeguarding online privacy. Data anonymization, for instance, includes removing or modifying identifiable information from datasets to protect individuals’ identities. Encryption secures data by converting it into a code that can only be decoded by authorized parties. Synthetic data generation creates artificial data that closely resembles real data but doesn’t contain any identifiable information. Differential privacy adds a small amount of controlled noise to protect sensitive information. Throughout our exploration, we not only explain the principles and techniques behind these technologies but also the tools used for each of these techniques and evaluation criteria and also examine their practical applications. By understanding their strengths, limitations, and real-world implementations, we gain valuable insights into how they contribute to the broader goal of ensuring privacy in our digital world.
Hardware-Based Implementation of Target Tracking in Unmanned Aerial Vehicles (UAVs)
(IEEE, 2023) Chalapathi, G.S.S.
Unmanned Aerial Vehicles (UAVs) have gained sig-nificant attention in various fields, including surveillance, search and rescue, and monitoring applications. One important application for UAV s is target tracking, which requires detecting and tracking a specific object of interest in real time. This paper surveys work done so far in the area of target-tracking in UAV s. It then presents a comprehensive hardware-based framework for target tracking in UAV s. This work utilizes the State-of-the-Art YOLOv8 (You Only Look Once) algorithm for target detection, an efficient high-speed target tracking model, and a Proportional Derivative (PD) control algorithm for precise drone movement control. YOLOv8 provides fast, accurate, and real-time detection of the object of interest, allowing the UAV to detect and identify the target object quickly and reliably. Subsequently, a robust tracking algorithm tracks the identified object across consecutive frames, ensuring accurate localization and trajectory estimation. Furthermore, a PD control algorithm is integrated into the system to enable precise and smooth drone movement. The proposed framework is integrated and used for target tracking in UAV s. Further, this framework is implemented on a UAV. The results demonstrate the effectiveness and robustness of the proposed framework, showcasing its potential for real-world applications.
Singaporean Conversational English-Malay Code-Switching Speech: An Analysis Based on Code-switching Points and Part -of-Speech
(IEEE, 2023) Chalapathi, G.S.S.
This paper investigates various code-switching prop-erties of conversational speech from bilingual English-Malay Singaporean speakers with data obtained from the National Speech Corpus (NSC) and provides baseline language models for various combinations between English-Malay monolingual and codeswitching transcripts. Specifically, the study analyzed the correlation between code-switching patterns and (i) trigger words and code-switched word pairs at code-switching points, and (ii) wordwise POS and pairwise POS tags. Our analysis shows there is a certain set of words that frequently “triggered” code-switching behavior, and speakers tend to code-switch more frequently around nouns. Additionally, we provide perplexities for language models built on the selected datasets. These perplexities could serve as baselines for future language models for Singaporean speech, especially, English-Malay code-switch speech.
Non-Fungible Tokens (NFTs)—Survey of Current Applications, Evolution, and Future Directions
(IEEE, 2023-12) Chalapathi, G.S.S.
Non-fungible tokens (NFTs) have become an exciting technology that provides a fresh perspective on asset ownership, provenance, and value exchange. NFTs, a blockchain-based technology, are distinct and indivisible cryptographic tokens used to confirm and record the ownership of digital and physical assets in an immutable and transparent way. The fundamental block of NFT is a smart contract built on a blockchain network. This contract contains specific information about the asset it represents, such as its unique identifier, metadata, and ownership details. The information is kept private and tamper-proof due to the decentralized and distributed structure of the blockchain, boosting faith in the token’s authenticity. The NFT is gaining popularity, but it is still in the developing stage. There is a need for a comprehensive survey to guide future research and development in NFTs. Thus, this paper presents the technical components of NFTs, their features, and the minting process. Further, this survey paper describes different token standards for NFTs. It presents various applications of NFTs in healthcare, supply chain, gaming, identity verification, agriculture, intellectual property, smart cities, charity and donation, and education. The article also emphasizes the significant difficulties faced currently in implementing NFT technology from the viewpoints of ownership, governance, and property rights, as well as security, privacy, and environmental effects. This work also elucidates the future directions to overcome the challenges in adopting NFTs in various applications.
LoRa-Based Wireless Sensor Network Testbed for Precision Agriculture Application
(IEEE, 2024) Chalapathi, G.S.S.
Over the past few years, Wireless Sensor Network (WSN) has seen many improvements, and there are various applications of WSNs in various domains. Most communication technologies have a trade-off between distance and power consumption, i.e., to reach a longer distance, high power is consumed. LoRa overcomes this problem by consuming less power and transmitting small data packets for a long distance. The objective of this research work is to use LoRa technology in Precision Agriculture applications to help the stakeholders in better decision-making. A small experimental testbed is set up for precision agriculture applications. This testbed had sensors to monitor pH, soil moisture, soil temperature, and NPK parameters. An Automatic Weather Station (AWS) is set up to monitor ambient weather parameters-temperature, humidity, rainfall, wind speed and direction, barometric pressure, solar radiation, and leaf wetness. These sensor parameters were collected at the LoRa Gateway and forwarded to a network server hosting the The Things Network (TTN) LoRa stack. Transmission statistics are collected and analyzed for this application for remote monitoring of agricultural farms for quick and efficient decision-making.
Multi-Objective MDP-Based Routing in UAV Networks for Search-Based Operations
(IEEE, 2024-05) Chalapathi, G.S.S.; Chamola, Vinay
Unmanned aerial vehicle (UAV) systems have gained widespread recognition due to their versatility and autonomy. Their deployment for disaster mitigation and management operations is seen as one of their most important applications over the past decade. In such UAV networks, routing plays a crucial role in determining network performance parameters such as network lifetime, data transmission latency, and packet delivery ratio. This paper presents a novel routing mechanism - Multi-Objective Markov Decision Based Routing (MOBMDP) for UAV networks carrying out search-based operations. MOBMDP models routing decisions in a Markov Decision Process (MDP) framework and uses Q-learning to take decisions. It compares routing paths using three metrics, viz., Remaining Energy of the Minimum Energy Node (REMEN), Power Distance ratio (PD), and Expected Delay (ED). Amongst these metrics, PD is a novel metric proposed by this work. PD simultaneously optimizes the energy efficiency and energy distribution in the network. Further, MOBMDP uses a novel reinforcement learning inspired method to estimate transmission delay in a given path. Intensive simulation studies compare MOBMDP to four state-of-the-art routing protocols. Results show a significant improvement in network lifetime, packet delivery ratio, energy efficiency, average data transmission delay, and error in delay estimation.
Privacy Utility Tradeoff Between PETs: Differential Privacy and Synthetic Data
(IEEE, 2024-11) Chalapathi, G.S.S.
Data privacy is a critical concern in the digital age. This problem has compounded with the evolution and increased adoption of machine learning (ML), which has necessitated balancing the security of sensitive information with model utility. Traditional data privacy techniques, such as differential privacy and anonymization, focus on protecting data at rest and in transit but often fail to maintain high utility for machine learning models due to their impact on data accuracy. In this article, we explore the use of synthetic data as a privacy-preserving method that can effectively balance data privacy and utility. Synthetic data is generated to replicate the statistical properties of the original dataset while obscuring identifying details, offering enhanced privacy guarantees. We evaluate the performance of synthetic data against differentially private and anonymized data in terms of prediction accuracy across various settings—different learning rates, network architectures, and datasets from various domains. Our findings demonstrate that synthetic data maintains higher utility (prediction accuracy) than differentially private and anonymized data. The study underscores the potential of synthetic data as a robust privacy-enhancing technology (PET) capable of preserving both privacy and data utility in machine learning environments.
EFTA: An Energy-efficient, Fault-Tolerant, and Area-optimized UAV Placement Scheme for Search Operations
(IEEE, 2022) Chalapathi, G.S.S.
Unmanned aerial vehicle (UAV) networks have widespread applications, ranging from surveillance and disaster management in the military domain to transportation and delivery of goods in the civilian domain. Regardless of the application, the placement of routing UAV nodes (especially in networks spanning long distances) is crucial in determining network performance parameters such as network lifetime and data transmission delay. In this paper, an Energy-efficient, Fault-Tolerant, and Area-optimized UAV placement scheme (EFTA) is proposed for search operations. A cluster-based UAV network is considered, in which the Cluster Members (CMs) are mobile and scan the geographic area of interest. The Cluster Heads (CHs) are quasi-static and route information from the CMs to the Ground Control Station (GCS). A multi-objective Cuckoo Search Algorithm is used to determine the placement of the CHs while minimizing energy consumption, maximizing area coverage, and maximizing tolerance to node failures. Further, a comprehensive analysis was performed against a state-of-theart UAV placement algorithm. The analysis showed that EFTA gives a significant performance improvement when compared to the competing placement scheme in fault tolerance, power consumption, network lifetime, end-to-end delay, and packet delivery ratio.

BITS Faculty Publications

Browse

Filters

Settings

Sort By

Results per page

Search Results