ICSSIT 2024: Papers with Abstracts

Abstract. Brain tumor segmentation is an essential step that is important for the diagnosis and treatment planning in healthcare. Brain MRI images are preprocessed in accordance with the suggested approach before data is gathered and ready for further analysis. The suggested study introduces a new strategy that uses the bio-inspired Particle Swarm Optimization (PSO) algorithm to segment brain tumor images. To improve accuracy and dependability, the segmentation model's parameters can be adjusted. Standard measures like Accuracy, Precision, Sensitivity, Jaccard index, Dice Coefficient, Specificity are used in performance evaluation to measure the effectiveness of the suggested PSO- based segmentation approach. The overall accuracy of the suggested method is 98.5%. Subsequent performance analyses yield better results of 91.95%, 87.01%, 92.36%, 90%, and 99.7% for Dice Score Coefficient, Jaccard Index, Precision, Sensitivity, and Specificity, respectively. Therefore, this method can be a useful tool for radiologists, supporting them in diagnosis of tumor in brain.

Abstract. The management of organ donation and transplantation systems is confronted with numerous challenges, spanning registration, donor-recipient matching, organ logistics, and ethical considerations. In response, this paper proposes a decentralized solution leveraging a private Ethereum blockchain framework. This paper establishes a secure, traceable, and privacy-preserving environment by utilizing smart contracts and a suite of innovative algorithms. This approach's robustness and trustworthiness will be highlighted by implementation details, testing procedures, and thorough evaluations that cover privacy, security, and confidentiality. Through comparative analyses, this solution emerges as a promising avenue for enhancing equity, efficiency, and patient confidence in organ donation and transplantation management.

Abstract. In the era where social media and technology intersect, the vast user base of social networks presents a challenge in handling massive data. The issue intensifies when making user suggestions amidst the overwhelming data flow. This analysis addresses the complexities arising from the abundance of data in social networks and proposes a solu- tion through advanced graph partitioning techniques, focusing on algorithms from promi-nent libraries like DGL and PyTorch. This analysis compares three graph partitioning algorithms for social network analysis: DGL METIS (edge-balanced and node-balanced), and PyG METIS. We analyze their performance on the Epinions social recommendation dataset, focusing on edge based and node based metrics and visualization of partitions.Our findings reveal: PYG METIS consistently exhibited suboptimal performance across various evaluation metrics, with the exception of achieving satisfactory results in node balance. Conversely, DGL Node Balanced METIS demonstrated marginally superior outcomes compared to DGL Edge Balanced METIS in terms of edge loss and average edges per partition and surpassed it in node balance.

Abstract. Globalization and advancements in trade facilitation have opened up new possibilities for polluting trade routes with counterfeits. Consequently, protecting goods and prod- ucts against counterfeiting has become critical for intellectual property owners worldwide. Yet, conventional safeguards of adding protective elements and safety features have proven inadequate. Over the past decade, blockchain technology has emerged with a promising po- tential to become a veritable tool in safeguarding supply chain authenticity. The proposed system presents an effective counterfeit detection system using blockchain technology. The proposed framework of the blockchain leverages a three-entity model - manufacturers, sell- ers (distributors), and consumers, ensuring transparency and immutability. The efficiency of the proposed system is demonstrated through experiments performed on various data.

Abstract. In recent times, suicide has emerged as a significant global public health concern, causing profound harm to individuals, families, and communities. Despite intensified efforts to address mental health challenges, suicide rates continue to surge in numerous regions worldwide. To effectively tackle this escalating problem, there is an urgent requirement for a thorough examination and understanding of the factors driving suicide occurrences. The primary objective is to conduct a comprehensive analysis of suicide rates utilizing advanced analytics techniques. This analysis endeavors to uncover the intricate network of socio-economic, demographic, and psychological factors shaping suicide trends across various countries and regions. Through meticulous investigation and interpretation of these factors, this research aims to gain insights that can inform targeted interventions and policies aimed at mitigating the prevalence of suicide and promoting mental well-being on a global scale.

Abstract. Laptops have become an indispensable part of everyone’s lives. They are beneficial for quickening the pace when it comes to any task. They serve the purpose of communication through online interactions and streamline other operations. The different kinds of laptops on the market have varying features designed precisely for your day to day needs. We have proposed a laptop price predictor that aims to predict the price of laptops. The results of the current study aid in understanding and identifying the factors that customers consider vital when purchasing a laptop. Further, the relationships amongst the important criteria were also identified. DEMATEL(Decision making trial and evaluation laboratory) method is employed to analyse the relationship between the variables and identify the most important factors. Mutual information regression (MIR) method was employed for feature selection and finally the selected features were input to the various Machine Learning and Deep learning algorithms for price prediction. Our study revealed that LightGBM Regressor and XGB Regressor performed the best, with the highest r2 scores of 0.81 after hyperparameter tuning.

Abstract. The proposed Virtual Assistant System (VAS) represents a new solution revolutionizing user interactions with technology by seamlessly integrating Natural Language Processing (NLP) and Artificial Intelligence (AI), facilitating effortless communication through voice commands. By employing cutting-edge speech recognition algorithms, the proposed system accurately translates the voice input into text, adapting responses based on individual user preferences over time. The proposed system offers a diverse range of functionalities including information retrieval, task automation, and smart home control to assist users in managing the tasks hands-free with an intelligent interface providing varying levels of technical expertise. Safeguarding user privacy and control, the system allows users to opt-in or opt-out of data collection with complete transparency and robust security measures. Continuous improvement through extensive testing and user feedback addresses the challenges like accurately interpreting complex commands, positioning the Virtual Assistant System (VAS) as a sophisticated, personalized, and privacy-aware solution at the forefront of virtual assistant technology.

Abstract. Human Activity Recognition (HAR) from video signals is increasingly crucial for surveillance, healthcare, robotics, and augmented reality applications. Accurately iden- tifying human actions is vital in our data-driven world, posing a significant technological challenge. This study introduces a comprehensive methodology for HAR, starting with the preprocessing of video frames using context-awareness. The context-aware frames are then fed into a two-stream framework, extracting spatial and temporal features in a complemen- tary manner. The spatial stream analyzes visual features from individual video frames, while the temporal stream focuses on dynamic aspects, capturing intricate motion patterns. This separation allows for a detailed analysis of video data, aligning with human perception of activities. The subsequent stage involves a late binding mechanism, enabling optimal interaction between spatial and temporal streams. Integration in a dense layer allows the model to harness interactions between these information streams, significantly improving recognition accuracy. Rigorous experimental validation confirms the efficacy and reliabil- ity of the proposed approach in diverse scenarios using real-world datasets HMDB51 and UCF50. The results demonstrate high accuracy, precision, recall, and F-measure for the combined spatial and temporal model compared to individual streams. This research con- tributes to advancing HAR technology, improving how computers interpret and recognize human activities in videos for practical and beneficial applications.

Abstract. Alzheimer's disease, a neurodegenerative disorder with profound societal impact, necessitates robust predictive models for early detection. This research delves into the realm of Alzheimer's prediction, concentrating on the efficacy of various machine learning algorithms. Gaining knowledge from previously published works in the Scopus database, we carried out a thorough review to identify recurring machine learning concepts. Our study synthesized information from diverse sources, revealing seven frequently employed machine learning algorithms. Through meticulous analysis of these algorithms, we discovered that Support Vector Machines (SVM) emerged as the most effective predictor, exhibiting superior performance in comparison to other models. The evaluation process included considerations of accuracy, sensitivity, and specificity, with SVM consistently outperforming its counterparts. Additionally, Random Forest emerged as a noteworthy alternative, showcasing commendable predictive capabilities. This study not only demonstrates the significance of machine learning in Alzheimer's prediction but also offers perceptive data for choosing the most efficient algorithmic approach. Our findings underscore the potential of SVM and Random Forest in enhancing diagnostic accuracy, laying the foundation for future advancements in early Alzheimer's detection and intervention. As the prevalence of Alzheimer's continues to rise, our work seeks to inform and guide scholars and professionals involved in the creation of efficient and reliable predictive models for improved patient outcomes.

Abstract. The forthcoming age of communication systems requires increased security, faster processing, and simpler encryption techniques. Quantum information and communication technology offer a revolutionary era featuring high-speed and inherently secure networks. With the looming prospect of quantum supremacy, traditional encryption systems face potential obsolescence in terms of security. In response to this challenge, quantum key distribution (QKD) emerges as an innovative solution for exchanging secret keys in a quantum based manner, addressing the limitations of conventional encryption methods. Within the proposed work, the focus extends to the implementation of a MIMO- QKD scheme tailored for terahertz (THz) frequency applications. The paper provides a comprehensive examination of both single-antenna QKD schemes and the performance of the MIMO QKD scheme. Simulation results underscore the indispensability of multiple antennas to mitigate the substantial free-space path loss encountered at THz frequencies.

Abstract. Microstrip antennas have gained significant attention in modern communication systems due to their compact size, ease of integration, and versatility. At the outset this paper explores the antennas operational parameters, such as frequency of operation, bandwidth, and gain. Utilizing the HFSS software, the antennas geometry, including the patch dimensions and feed structure, is optimized to meet these requirements. Designing an antenna to attain specific requirements such as compact size, high- performance in operation tailored to specific frequency and bandwidth require ments is a critical job an antenna engineers. Electromagnetic simulations are conducted to analyze various antenna characteristics, including radiation pattern, impedance matching and gain. Through iterative designs, simulation and optimization processes. The proposed antenna's performance is being assessed across multiple antenna parameters, encompassing Return Loss, Bandwidth, Gain, and S-Parameters. Respective antenna design modifications are implemented to fine-tune the antennas performance. Using HFSS software, the paper increases understanding of microstrip patch antenna design. It also offers practical guidance for creating high-performance antennas suitable for diverse applications like wireless communication, satellite communication, and radar systems.

Abstract. AI has emerged as a transformative force for startups across various industries. It offers automation, data-driven decision-making, personalization, and predictive analytics, enabling startups to improve efficiency, gain a competitive advantage, and scale their operations. The startup resource dashboard application uses Machine learning to scrape data from a variety of sources, such as government websites, industry publications, and social media. The application extends beyond simple data gathering by incorporating machine learning to craft an advanced matching algorithm. It uses Large Language Models trained exclusively on data relevant to the current economic information and other data that helps the startups to take data driven decision. Furthermore, vector embeddings are employed to enhance the context and relevance of the generated responses. Encoding the words and phrases in the High-dimensional vectors, the model gains a better undersign of the startup eco-system facilitating more accurate and insightful recommendations. By bridging the gap between Advanced AI technologies and specific needs of startups, this methodology improves the innovation and success within the startup environment.

Abstract. This research assesses sentiment analysis on Twitter posts about Twitter tweets, comparing ma- chine learning techniques. Four different models are used to classify and their efficacy is evaluated. Addressing balanced datasets, SVM excels in balanced sentiment scenarios. Using metrics like accuracy and recall, the study offersinsights for decision-making in marketing and social studies. Emphasizing machine learning effectiveness, the research suggests improvements for sentiment analysis in diverse domains, particularly in understanding positive and negative Twitter tweets.

Abstract. A vital industry, agriculture is essential to maintaining the world's food security. Accurate and timely information on crop yields is essential for farmers, policymakers, and stakeholders to make well-informed choices concerning cultivation, dissemination, and allocation of resources in crop management. Traditional methods of data collection and analysis in agriculture have limitations in terms of efficiency and accuracy. This paper presents a Crop Yield Information System (CYIS) that leverages Market Basket Analysis (MBA) techniques to improve agricultural yield forecasting and offer insightful information for stakeholders in the agriculture sector.
The CYIS is designed to collect and process a vast amount of agricultural data, including historical crop yield records, meteorological data, soil quality information, and market price data. Market Basket Analysis, a well-established data mining technique in retail, is adapted to identify patterns and relationships among various agricultural variables. The application of an MBA in agriculture allows the system to uncover hidden associations between factors such as weather conditions, crop types, and yield outcomes.
Key features of the CYIS include data collection through IoT devices, data pre- processing for quality assurance, and the application of association rule mining algorithms to derive meaningful insights. The system offers a user-friendly interface,enabling farmers, agronomists, and policymakers to access real-time yield predictions,make informed decisions, and implement precision agriculture practices. Furthermore,the CYIS integrates market prices and demand data to provide stakeholders with market insights, facilitating better decision-making related to crop production, storage, and distribution.
This study adds value to the agricultural domain by offering a strong and innovative approach to crop yield prediction and decision support. The CYIS, utilizing Market Basket Analysis, offers a data-driven solution to optimize crop production, reduce resource wastage, and enhance food security. By harnessing the power of data analytics, this system empowers agriculture stakeholders to respond to changing conditions, market dynamics, and climate challenges with more accuracy and efficiency. The Crop Yield Information System using Market Basket Analysis represents a significant step toward enhancing agricultural sustainability and food security. This research opens new avenues for leveraging data analytics in agriculture, and its implementation has the potential to revolutionize how farmers, agronomists, and policymakers approach crop production and distribution.

Abstract. Speech is utilized in human-machine connection and serves as a signal of human involvement. The Speech Emotion Recognition (SER) system is a novel form of this interactive system. Sufficient intelligence is provided by the SER to facilitate effective human-computer interaction. Based on the speaker's words, the SER system classifies emotions into groups such as "neutral," "calm," "happy," "sad," "angry," "fearful," "disgust," and "surprise." Languages and machine learning models suitable for SER are defined in this paper. Deep learning is used by this system to effectively classify and learn from multidimensional data. Primary results for a system using the LSTM algorithm and MFCC feature tools are also presented in this work. For the simplicity of user engagement, we have then implemented this model as a website through the usage of a third party.

Abstract. Kidney abnormality is one of the major concerns in modern society, affecting millions of people worldwide. To diagnose different kidney abnormalities, a narrow-beam X-ray imaging procedure called computed tomography is commonly used, creating cross- sectional slices of the kidney. Deep learning models have shown promise in classifying and segmenting kidney abnormalities from these CT images. However, these models often present challenges for clinicians in interpreting their decisions, leading to a "black box" system. In response to this issue, this study proposes a transfer learning technique for the detection of kidney cysts, stones, and tumors. With improved results and enhanced interpretive power, the proposed work empowers clinicians with conclusive and understandable outcomes. ⁠Furthermore, this research delves into the selection of the optimal optimizer for kidney abnormality prediction. The study performs a comprehensive performance evaluation of three popular optimizers - Adam, Adadelta, and AdamW - on coronal and axial abdominal and program images. The findings shed light on the most suitable optimizer, ensuring better training and generalization of the transfer learning model. Finally Leveraging the Flask framework, the developed web application enables seamless and efficient prediction of kidney abnormalities, potentially revolutionizing the clinical decision-making process and enhancing patient care.

Abstract. The Integrated Remote Power Control System (IRPCS) represents a breakthrough in intelligent energy management by leveraging Internet of Things (IoT) technology to enhance power efficiency and sustainability. This innovative system utilizes a combination of Arduino microcontrollers, Wi-Fi modules, and advanced sensors, including PIR for motion detection and ambient light sensors, to monitor and optimize electricity usage across various environments. By enabling real-time control and monitoring of electrical devices by the user, IRPCS significantly reduces energy wastage and operational costs, while promoting environmental conservation. Its scalable and adaptable framework ensures it is suitable for both residential and industrial applications, offering a versatile solution to the modern challenges of energy management. IRPCS stands at the forefront of energy innovation, embodying the potential of IoT to transform power consumption patterns towards a more sustainable and efficient future.

Abstract. Customer segmentation is a critical component of any marketing strategy for contemporary companies that are ready to compete in a market that is fiercely competitive today. The goal is to use segmentation of customers methodologies to better decision-making, marketing tactics, and customer satisfaction levels in general. The project will get started by gathering and studying various client data, including comments, purchase history, online activity, and demographics. The consumer base will be segmented into various segments based on shared traits and preferences using data analytics as well as machine learning algorithms. The primary objective is to leverage these segments to optimize various facets of business operations, such as marketing campaigns, product development, and customer support. Enterprises can position themselves for enduring expansion and triumph in the constantly evolving and fiercely competitive commercial sphere by adeptly segmenting their consumer demographics and adjusting their approaches accordingly. Through these insights, businesses can achieve more efficient resource utilization and improved ROI (Return On Investment). Emphasizing the significance of customer segmentation as a strategic tool enhances business performance.

Abstract. Ensuring maritime security and surveillance demands advanced technological solutions, and satellite imagery has emerged as a pivotal asset in this domain. This paper introduces an innovative approach for ship detection in satellite imagery, integrating convolutional neural networks (CNNs) and artificial neural networks (ANNs). The amalgamation of these neural network architectures aims to overcome the intricate challenges associated with maritime surveillance, including dynamic environmental conditions and the diverse nature of vessels. The Convolutional Neural Network (CNN) component is used for extracting complex spatial features from satellite imagery, allowing for the identification of potential ship-related patterns. Acting as a specialized detector, the CNN navigates the complexities of maritime landscapes, discerning vessels from varying backgrounds and environmental factors. Complementing the CNN, the Artificial Neural Network (ANN) component refines the high-level features extracted, facilitating advanced analysis and reducing false positives. The synergy between CNN and ANN contributes to a comprehensive ship detection system, enhancing accuracy and adaptability in real-world scenarios. Extensive experimentation on diverse satellite imagery datasets validates the effectiveness of the proposed integrated approach. The results demonstrate a high performance compared to individual neural network models, ensuring the system's resilience to the changing conditions. The versatility of this integrated solution positions it as a valuable asset in real-time maritime surveillance, promising to increase the standards of maritime security and surveillance operations.

Abstract. The study looks at how crucial it is to have individualized visitor information systems in order to maximize the travel industry's economic potential. The research findings introduces a novel hybrid method that combines content- based algorithms with collaborative filtering algorithms to provide tourists accurate and personalized recommendations. To make these recommendations more accurate, item comparison using TF-IDF and cosine similarity is utilized. To ensure that it is applicable to a broad audience, the research expands its scope to include data on tourists from both India and throughout the world. In order to improve the entire user experience, an intuitive interface is developed and visual material, such as photographs, is integrated. It can be done using a user- centric approach. The research advances the tourism business by focusing on efficiency and relevancy, which benefits travelers as well as the industry as a whole.

Abstract. Computer vision is a rapidly advancing field with profound implications for the healthcare industry. This technology leverages the power of artificial intelligence and image processing to extract valuable insights from medical images and videos. So in my project breast cancer is taken as priority. In contemporary healthcare, the timely and accurate detection of breast cancer is paramount to improving patient outcomes and reducing mortality rates. Breast Ultrasound Imaging (BUSI) stands as a pivotal non- invasive tool in this endeavor. However, to unlock its full potential, advanced image processing techniques are imperative. OpenCV, a versatile computer vision library, plays a critical role in this context and specifically tailored for analyzing BUSI ultrasound images. The methodology encompasses key stages. A meticulously annotated datasets of BUSI ultrasound images, discerning the presence or absence of breast cancer, is curated. Leveraging OpenCV capabilities, the datasets undergoes pre- processing, including re-sizing, normalization, and enhancement, to optimize an image quality for subsequent analysis. MobileNetV2, selected for its computational efficiency, serves as the foundation for transfer learning, synergistically integrated with OpenCV for robust feature extraction. The trained MobileNetV2 model, enriched by OpenCV image processing capabilities, undergoes rigorous evaluation on an independent test set, employing various performance metrics. This assessment aims to quantify the model's proficiency in breast cancer detection. The amalgamation of OpenCV and MobileNetV2 with BUSI ultrasound images seeks to achieve accurate and reliable results, underscoring the critical role of advanced image processing in modern healthcare. The developed model demonstrates potential for real-world deployment, particularly in web-based systems, enabling healthcare professionals to detect breast cancer early. Users can seamlessly upload BUSI breast ultrasound images for analysis, with the model providing predictions regarding the presence or absence of breast cancer. This integrated approach not only enhances diagnostic accuracy but also expedites patient care, exemplifying the indispensable role of OpenCV in modern healthcare applications.

Abstract. Agriculture, humanity's foundational activity, encompasses a rich tapestry of practices essential for sustaining life on our planet. From the cultivation of crops and the rearing of livestock to the management of forests and fisheries, agriculture stands as the bedrock of food production, economic development, and environmental stewardship. Across cultures and civilizations, agriculture has played a pivotal role in shaping societies, landscapes, and livelihoods, fostering connections between people and the land they cultivate. In the contemporary world, agriculture faces a myriad of challenges, from population growth and climate change to resource scarcity and ecology degradation. In light of the growing global population, surpassing 8.1 billion people, and climate change alters weather patterns and exacerbates extreme weather events, the pressure on agricultural systems to produce more food while conserving natural resources and mitigating environmental impact has never been more acute. Through the analysis of datasets sourced from the Government of India, encompassing critical factors such as pH, temperature, rainfall, humidity, and NPK substance, the study aims to provide actionable insights for agricultural stakeholders. At the heart of the project lies the recognition that informed decision-making is essential for driving efficiency, resilience, and profitability in agricultural operations. By empowering stakeholders with data-driven predictions and informed strategies, the project aims to enhance agricultural business development, enabling farmers, policymakers, and other stakeholders to make more informed choices. Whether it's optimizing crop selection, improving resource allocation, or mitigating risks associated with climate variability, the project endeavors to provide tools to the stakeholders and managing the complexities of modern life requires knowledge and experience agriculture. The predicted accuracy is 99.69%.

Abstract. Deepfake content is created or changed artificially utilizing AI strategies to make it genuine. This research addresses the evolving challenge of detecting deepfake audio content, as recent advancements in deepfake technology have rendered it increasingly challenging to distinguish fabricated content. Leveraging machine and deep learning methodologies, specifically employing Mel-frequency cepstral coefficients (MFCCs) for sound component extraction, we focus on the Genuine-or-Fake dataset — a cutting-edge benchmark dataset generated through a text- to-speech (TTS) model. This dataset is arranged into sub-datasets because of sound length and spot rate. This study reveals that the Convolutional Neural Network (CNN) models exhibit the highest accuracy in identifying deepfake audio within the for-rerec and for-2-sec datasets. Meanwhile, the gradient boosting model performs well in the for-norm dataset. This study illustrates the CNN model's outstanding performance on the for-original dataset, outperforming other cutting-edge models. This study advances the field of deepfake recognition, especially in the areas of audio manipulation, demonstrating the efficacy of CNN models in detecting fake content.

Abstract. Creating a microstrip antenna for Wireless Body Area Network (WBAN) applications is a major challenge in the Internet of Things (IoT). Sports monitoring, healthcare, and other industries are among the areas for which this antenna is intended for usage. It operates in the 9.4 to 18.3 GHz UWB frequency range. It follows a meticulous modelling and design process. To miniaturize without compromising performance, the goal is to employ cutting- edge techniques like compact architectures and fractal geometries. In order to provide dependable connectivity in body-centric Internet of Things applications, the antenna aims to demonstrate perfect impedance matching and an omnidirectional radiation pattern.

Abstract. Power reduction is one of the major development challenges of this period. In high performance digital frameworks, such as microchips, DSPs, and other applications, low power circuit designs are a requested feature. The primary factors taken into account when comparing any circuit or design are power and speed. When proposing any new design, designers should keep in mind the very impressive but declining chip area. 2x4 Decoder Binary inputs are converted to associated output bits in a pattern using 12 T. A new 2x4 decoder In this work, area optimization with the use of 12T is suggested. The 2:4 decoder is also executed using CMOS logic. The new design and CMOS logic are evaluated in terms of delay and power. Compared to CMOS logic architecture, which typically operates at 0.8V, the new 2x4 decoder design is optimized for power at a rate of 60.72%. PMOS of width (Wp)=64n& length(Ln)=16n,NMOS of width (Wn)=32n & length (Ln)=16n,Fan-out=1, and Wp/Wn=1, after that the designs are assessed at 0.8V with 16nm Mentor Tanner Tool has been used to validate the suggested procedure.

Abstract. This literature review aims to navigate the vast landscape of image captioning, an interdisciplinary field that lies combining natural language processing and computer vision. We start with a detailed examination of the CNN-to-Bi-CARU model, an attention-based bidirectional architecture for comprehensive contextual information extraction. The application of this model in image captioning therefore necessitates detecting image features and objects, and identifying them precisely. Attention mechanisms are important for securing precise matching regarding changes in focused content during caption generation. The efficiency concerns have been highlighted by the CNN-to-Bi-CARU model that has taken less time in coming up with images during inference. Stability is acknowledged even as improvements are proposed for a perfect BDR-GRU system. The experimental phase investigates different loss functions and optimizers leading to selecting cross-entropy as a loss function and Adam optimizer to achieve BLEU-4 metrics and better accuracy. The introduction of a new framework allows for the estimation of significant regions in images. The approach relies on image captioning, which incorporates semantic information while estimating important regions on basis of subject and object words contained in those captions. Experimental results confirm that the technique can estimate important regions with sensitivity rivaling human perception. In regard to remote sensing image captioning, this exploration ends up with an encoder-decoder model. Instead of traditional token generation, the model supports continuous output representations, using a proposed loss function to optimize semantic similarity at sequence level. This novel way may have a great impact on language generation in the context of remote sensing imagery. Viewing the diverse methods that were explored, problems that have been identified and inventions that have been realized, this paper provides an overview of the landscape and a call for further research. The importance of stability and loss functions in this emerging area emphasizes it’s dynamic nature, which portends improved image captioning. In conclusion, the present proposal presents an overview on what the field is currently experiencing thus serving as a basis for more improvement and exploration in image captioning which is considered fascinating.

Abstract. Due to its potential in many industries, including law enforcement, facial recognition technology has attracted a lot of interest recently. The proposed face recognition method compares the facial traits of unidentified people with a database of known missing people using high-resolution images and advanced algorithms. The first step in the facial recognition procedure is to use specialized cameras to take clear photographs of the unknown person's face. Advanced computer vision algorithms are then used to process these photographs in order to extract distinctive facial characteristics such as the separation between the eyes, shape of the nose, and contours of the face. Using pattern recognition tools, these attributes are then compared with a database of people, who have gone missing in the past. The proposed system uses machine learning algorithms to enhance its accuracy and dependability over time. It can adapt to changes in stance, lighting, and age progression, enabling it to recognize missing people even after a sizable amount of time has passed. The efficiency of the proposed identification process can be improved further by integrating this system with the databases already used by law enforcement, which will ultimately improve the possibilities of finding missing people and reuniting them with their family and loved ones.

Abstract. The development of Machine Learning (ML) and DevOps, often referred to as MLOps, has revolutionized the healthcare sector by offering efficient and scalable solutions for disease classification. Machine Learning models, particularly deep learning algorithms, have demonstrate d remarkable performance in classifying diseases from various medical data modalities such as images, genomic sequences, electronic health records, and more.

Abstract. India's Public Distribution System (PDS) is a government policy that distributes commodities to needy people at fixed rates. However, manual intervention in weighing materials leads to inaccurate measurements and illegal use of consumer materials. At present, we have a system where the products are manually forwarded to the consumer by scanning their ration card and by verifying their fingerprint. But this was not enough for the corruption to stop. So, a system has been proposed where there will be a two-step verification as usual. But, when the step to handle the commodity comes, we use automation instead of manual labor. Consumers are provided with an RFID card that acts as a ration card with a unique identification number. The RFID card is scanned by an RFID reader interfaced with a microcontroller at a ration shop. The system then asks for the customer’s fingerprint, and it identifies whether the fingerprint matches the card. If both matches then the system automatically activates after verifying the consumer's allotted amount of rice, wheat, and sugar. The consumer receives the products when
they provide their bowl or cover at the bottom of the container. The motor at the entry point rotates, such that the selected item falls into the main outlet, and the consumer gets the material by weighing it on the load cell. A confirmation message is sent to the consumer's mobile number upon successful completion of the purchase. It is much more efficient than the manual system as it reduces labor costs and also stops the way to smuggle goods.

Abstract. Mineral exploration is vital for ensuring a reliable source of raw materials that are necessary for contemporary living and the shift towards environmentally friendly technologies. The mining process entails costly procedures aimed at detecting regions with inherent mineral concentrations in the Earth's crust. Combining artificial intelligence and remote sensing techniques has the capacity to greatly decrease the expenses linked to these operations. Here, it presents a strong and intelligent model for mineral exploration that is specifically designed to identify possible areas to extract the desired composition of mineral. Our approach incorporates cutting-edge developments in artificial intelligence and remote sensing, and introduce a sophisticated deep-learning process that utilises a random forest algorithm to examine the dataset. The main goal is the find out the type of ores to be extracted from the given minerals. This technique has a wider scope than just identifying things. It can also be used to find the type of soil to extract the ores. This versatile method is not restricted to single ores and can be utilized for different ore deposit models and dataset types. The incorporation of deep learning into the analysis of ores data is a groundbreaking progress in the domain of mineral exploration. It can improve the efficiency, precision, and cost-effectiveness of identifying areas with abundant minerals, making a significant contribution to the sustainable acquisition of raw materials and the worldwide shift towards environmental.

Abstract. This study has proposed a new transaction method for people, who don't trust one another to send
secure money by using a decentralized escrow technology. The proposed protocol takes an escrow,
a third-party smart contract to receive tokens before a transaction get completed. The tokens will
be released by the escrow upon fulfillment of the payment requirements. Delivering the agreedupon commodity or service and paying the associated payment are the obligations of each party to
perform a transaction. It should not be possible for one side to pull out a bargain at the cost of the
other. The Escrow may be provided with the necessary information by using the oracle pattern if
the conditions of payment are contingent on external data, such as the time of product shipment.
Once a smart contract's code is recorded on the blockchain, it cannot be changed. This guarantees
the safety of the escrow feature.

Abstract. In a dynamic contemporary business landscape sustained longevity of startups depend on the financial performance. While identifying the factors that add value to financial performance, sustainable business practices have emerged as a cornerstone. Integration of socially and environmentally responsible practices aligns the startup with competitive and financial viability. This study analyzes the intricate relationship between sustainable business practices (SBP)and competitive advantage (CA). The synergistic influence of SBP and CA is anticipated to intricately shape and define the Financial Performance (FP) of new enterprise. The study employs a multiple regression analysis to investigate multifaceted impact of sustainability on economic viability of emerging ventures. The study seeks to unveil the complex dynamics governing this tripartite relationship thereby paving way for strategic imperatives that propel startups towards a more resilient and sustainable future.

Abstract. In the dynamic landscape of the entrepreneurial ecosystem, where ventures are adopting sustainable business models, it becomes essential to analyze whether the adoption of sustainable business practices has a competitive advantage. This study investigates the complicated relationship between Sustainable Business Practices (SBP) and Competitive Advantage (CA) on a robust dataset of diverse startups and applies advanced regression analysis with a 10-fold cross- validation methodology. The results show a strong and positive association between SBP and CA. The results not only establish a statistical relation but also highlight the significant impact of sustainable practices on startup competitive advantage. This research contributes to both academic literature and management decision-making by giving detailed insights, as well as a strategic roadmap for entrepreneurs looking to exploit sustainability as a cornerstone of their competitive strategy.

Abstract. This paper introduces a robust web application for secure image transmission, integrat- ing AES encryption and hashed data to safeguard both the confidentiality and integrity of shared images. Employing cyberblock chaining mode enhances the encryption process, fortifying data protection. The sender module employs AES encryption, while the receiver module decrypts and displays the images securely. User authentication ensures the integrity of data transfer, while a user-friendly inter- face streamlines the image transmission pro- cess. This application finds relevance across diverse sectors including healthcare, legal, and corporate environments where image confidentiality and data integrity are paramount. Index Terms—Image encryption, Advanced encryption standard, hashed data, secure image transmission, web application, Steganog- raphy, security, confidentiality.

Abstract. Crowdfunding is a transformative fundraising method, enabling the collection of small contributions from a large audience, often facilitated through online platforms. This approach, leveraging the collective support of a diverse crowd, offers individuals, entrepreneurs, and organizations an alternative to traditional financial intermediaries for financing their projects. Recently, crowdfunding has become a potent tool for funding innovative ventures, yet faces challenges related to transparency and security. In response, this system proposes a blockchain - based crowdfunding platform, employing smart contracts, decentralized identity, and a tokenized system to enhance transparency and trust. The platform's key features include smart contracts governing the crowdfunding process, a user-friendly interface for campaign creation with details on proposed system goals, and seamless integration with cryptocurrency payment gateways, ensuring a transparent and trustworthy ecosystem for both project creators and backers.

Abstract. Cardiovascular Diseases (CVD) are the most prevalent global health concern that demands p romp t at t ent ion given their substantial role in increasing mortality figures. Owing to the need for early detection to alleviate the inimical effects of CVD, this study makes extensive use of machine learning techniques including Support Vector Machine (SVM), AdaBoost, XGBoost, and Decision Tree in the early prediction of cardiovascular diseases. The robustness of the model will be enhanced by assessing three diverse datasets enriched with various types of patient information to derive the most efficient model. Through this study we conduct thorough performance evaluations, considering various evaluation metrics such as Accuracy, Sensitivity and False positive rate, aiming to identify the most effective machine learning model for early CVD detection. The results help shed light on important findings that can lead to improved outcomes, which help in the fight against cardiovascular diseases.

Abstract. The increasing amount of data generated by increasing road networks poses a significant challenge in the area of accident prediction. This investigation delves into the use of graph partitioning techniques to address the complexities associated with handling large graph datasets in the context of accident prediction. Graph partitioning techniques are critical in breaking down complex graph structures into manageable components, promoting parallelism, and enabling scalable computations. The computational burdenis distributed by strategically dividing the road network into sub-networks, resulting in faster analysis. This study investigates various graph partitioning algorithms and evalu- ates their effectiveness in maintaining the overall integrity of the road network during the partitioning process. Furthermore, using robust evaluation metrics, this study compre- hensively compares various graph partition methods, providing valuable information to choose the most effective strategy for a specific traffic network, thereby advancing robust and optimized solutions for accident prediction.

Abstract. This study focuses on enhancing the performance of the ResNet50 model on the Intel dataset, a collection of images depicting diverse natural scenes under various environmental conditions. While ResNet50 has shown remarkable performance in image classification tasks, its application to the Intel dataset reveals certain limitations in accurately discern- ing subtle features within scenes. To address this, proposed architectural modifications to ResNet50 aimed at capturing intricate features specific to the Intel dataset. Four distinct modifications are introduced, tailored to exploit different aspects of scene complexity present in the dataset. Through extensive experimentation and evaluation, we demonstrate the effectiveness of these modifications in improving the model’s classification accuracy on the Intel dataset. the findings not only contribute to advancing deep learning methodologies for image analysis but also underscore the importance of tailored model design for specific task domains.

Abstract. Ensuring children's safety is a top priority for parents globally amidst urban and suburban complexities. Traditional supervision methods are increasingly impractical in today's fast-paced society, leading to a rise in missing children cases, such as the 47,000 reports in India in 2023. To combat this, an innovative solution harnessing computer vision technology, specifically OpenCV, has been developed. This system offers real-time monitoring within designated safe zones, utilizing sophisticated image processing and object tracking.

Abstract. Utilizing an Arduino Uno can significantly enhance a RC car's capabilities, acting as its central control unit to manage tasks such as motor control, sensor integration, and remote control interpretation. Meanwhile, incorporating an ESP32 camera module into the RC car enables live video streaming from the onboard camera, allowing remote operation and providing an engaging experience. Leveraging the ESP32-CAM's functionalities, this research study has developed a novel vehicle monitoring and control system. Although the Arduino Uno's limited processing power and GPIO pins make it unsuitable for camera-related tasks, it can still activate the ESP32 camera module whenever required. This setup enables manual motor control via a smartphone app and real-time video surveillance, offering a comprehensive situational awareness to the driver. Additionally, a dedicated smartphone app facilitates direct communication with the developed system, granting manual control over the vehicle's motor driver, enhancing flexibility, and enabling quick responses in emergency situations. The integration of modern technologies enhances accident prevention control, elevating the overall user experience and safety of the RC car.

Abstract. The early detection of Acute Lymphoblastic Leukemia (ALL) poses a significant chal- lenge in the medical field due to the subtle morphological features of ALL cells, which often resemble healthy cells. This necessitates the expertise of experienced hematologists, a re- liance on human interpretation that introduces subjectivity and labor-intensive processes. Consequently, timely diagnosis and treatment initiation can be hindered. By leveraging the capabilities of machine learning,the paper aim to establish a system that can accurately distinguish between healthy and ALL cells, thereby reducing the reliance on subjective human interpretation and expediting the diagnostic process. This systematic review thor- oughly examines the use of deep learning for classifying and detecting acute leukemia. This study discusses many stages such as preprocessing, augmentation, segmentation, and feature extraction that are taken before classification. It also addresses the issues faced by the authors in different datasets. This research study examined and compared several benchmark models VGG16, VGG19, Inception, Xception, Efficient NetB0, ResNet50, and ResNet101. Out of these models, ResNet101 came as the top performer with a Validation Accuracy of 76.36%, Validation Precision of 75.85%, and Validation Recall of 76.36%. This comparative analysis aims to elucidate the strengths and weaknesses of these models, contributing valuable insights.

Abstract. Non-Alcoholic Fatty Liver Disease (NAFLD) is a prevalent liver condition that necessi- tates accurate and non-invasive diagnostic approaches for effective treatment. This research addresses the challenges associated with present invasive procedures, such as liver biopsies, and proposes a novel diagnostic tool. Inspired by the limitations of existing methods, our project focuses on revolutionizing routine check-ups for middle-aged individuals at risk of NAFLD. Instead of traditional invasive biopsies, our diagnostic tool recommends a blood test, ensuring accurate identification and timely intervention. The conventional diagnostic methods for NAFLD involve imaging and invasive procedures, leading to accessibility and accuracy issues. In response, our user-friendly web application utilizes standard blood test findings to provide a quick and painless NAFLD diagnosis. This approach aims to create an affordable, easily accessible tool that minimizes patient discomfort. Leveraging a dataset of 3,237 individuals from NHANES III, our model achieves an outstanding accuracy rate of 89%. The dataset includes both NAFLD-positive and NAFLD-negative cases, ensur- ing a robust and representative model. In summary, this work makes significant strides in developing a blood-based, non-invasive method that enhances accessibility to NAFLD diagnostics through a user-friendly web application. The proposed tool offers a convenient option for patients and equips healthcare providers with an effective NAFLD diagnostic tool, fostering better patient care outcomes through early detection and intervention.