Browse
Recent Submissions
Item Facial Traits Using Neural Networks(NHCE, 2022) C HARSHA VARDHAN REDDY; S YERRINATHA REDDY: B YASWANTHThis paper presents a facial expression recognition system employing Bezier curves approximation technique. The system is based on facial features extraction using the knowledge of the face geometry and approximated by 3rd order Bezier curves representing the relationship between the motion of features and changes of expressions. For face detection, color segmentation based on the novel idea of fuzzy classification has been employed that manipulates ambiguity in colors. Experimental results demonstrate that this method can recognize the facial expressions with an accuracy of more than 90cases. Finally, the system has been implemented using a manipulator robot and issuing facial expression commands. From human face structure, We divide in 3 regions such as right eye, left eye and mouth areas from the face image. Firstly comes the face detection and then detection of the skin region. We crop the facial skin region and connect the largest skin region to detect the skin surface of the human face.Item Perception of Flaws in Steel Using resunet Algorithm(NHCE, 2022) KIRAN RAJ SURESH MALALI: LOKESH B S: NANDAN KNDue to the limitations of science and technology, manual inspection methods are still used in traditional industrial manufacturing to detect product surface faults. There is much opportunity for improvement in the present, popular deep learning-based technique due to its low degree of accuracy. This technique is slow and unsuccessful because of the limitations of manual labour and out-of-date technologies. The goal of this thesis is to determine whether it is possible to automate this process using state-of- the-art computer technology and picture fault categorization using one of the most current deep learning approaches. As a first step towards developing a fully automated model for pixel-wise segmentation that can identify steel faults in photos, we advise ResUNet, a hybrid of U-net and ResNet50 architecture. Our experimental results show that the suggested architecture produces good segmentation results on datasets that are made publicly available. Based on the results of controlled testing, the research concludes that it is possible to achieve a dice coefficient of more than 76%. Keywords: Surface defects, Deep learning, Convolution neural network (CNN), Image Classification, ResNet50.Item Healthly: Multiple Disease Prediction(NHCE, 2022) POTTURI DOLA SAI VARMA : POLEPALLE EEKSHITH KUMAR: POTTURI CHANDRASEKHAR RAJUHealthcare field. In this project, three popular classification models, namely Decision Tree, XGBoost, and Logistic Regression, were utilized to predict three different diseases: diabetes, heart attack, and Parkinson's disease. The use of machine learning algorithms to predict multiple diseases has been increasingly popular in the For diabetes prediction, the dataset used was from the Diabetes Diseases, consisting of observations and features. The decision tree model, XGBoost, and logistic regression achieved an accuracy. For heart attack prediction, the dataset used was from the Heart, consisting of observations and features. The decision tree model XGBoost achieved and logistic regression achieved an accuracyFor Parkinson's disease prediction, the dataset used was from the Parkinson, consisting of observations and features. The decision tree model XGBoost and logistic regression achieved an accuracy, Overall, the results indicate that XGBoost is the most effective model for predicting all three diseases. These findings suggest that machine learning algorithms can be useful in predicting multiple diseases and can provide valuable insights for healthcare professionals in making informed decisions about patient care. Keywords: Decision Tree, XGBoost, and Logistic Regression and Machine learning techniquesItem Detection of Distributed Denial of Service Attack in SDN(NHCE, 2022) K. JOHN PRASAD: K. VAMSI RAJU: MANJUNATH TUA network architecture known as software-defined network (SDN) allows for the virtual construction and design of hardware components. It is able to modify the settings of network connections in a dynamic manner. Because it is a fixed connection, it is impossible to change dynamically in the traditional network. Despite its usefulness, SDN is still susceptible to DDoS attacks. The internet is threatened by the DDoS attack. The machine learning algorithm can be used to stop the DDoS attack. Multiple collaborative systems are used in the DDoS attack to simultaneously target a specific server. In SDN, the control layer is in the middle, linking the application and infrastructure layers, where software controls devices in the infrastructure layer. In order to identify malicious traffic, we propose a Decision Tree and Advanced Support Vector Machine (ASVM) machine learning approach. The results of the test demonstrate that the Decision Tree and Advanced Support Vector Machine (ASVM) algorithm offers superior detectionrate and accuracy.Item Bitcoin Price Prediction(NHCE, 2022) TANUJ SINGH RATHORE: KRISHNA AGRAHARI: SAUMIK SHUBHAMThe new methods for tackling prediction issues in time series include machine and deep learning-based algorithms. It has been demonstrated that using these strategies yields more accurate findings than using traditional regression-based modelling. Long Short-Term Memory (LSTM), a type of artificial Recurrent Neural Network (RNN) with memory, has reportedly been shown to outperform Auto-regressive Integrated Moving Average (ARIMA) by a significant margin. Additional "gates" are incorporated into the LSTM-based models in order to memories longer input data sequences. The main query is whether the LSTM architecture's built-in gates already provide a strong prediction and whether further data training is required to further enhance the forecast. By traversing the input data twice (from left to right and from right to left), bidirectional LSTMs (Bi-LSTMs) offer extra training. If Bi-LSTM outperforms standard unidirectional LSTM after adding training capabilities, this raises the research topic of interest. This study compares and analyses the behaviour of the Bi-LSTM and LSTM models. The goal is to determine the extent to which additional layers of training data may be used to fine-tune the relevant parameters.Item Law Enforcement Companion(NHCE, 2022) DEEPESH KUMAR: SHREYAS CHAUDHARY: SOURABH V.N:Finding fugitive offenders after they have committed a crime or an illegal act takes time and effort. It is challenging for law enforcement authorities to complete this work on their own given the rising population density and the size of any nation's landmass. This cycle is both times and works seriously. In this paper, we tried to suggest a different framework for criminal Distinguishing & Recognition using Deep learning and Heroku Cloud, i.e., Cloud Computing, which, assuming it is used by our Crime Control Organizations, would help them catch criminals from CCTV images or images uploaded by the public if seen anywhere. This system is in place to assist in capturing criminals and anyone who can upload information indicating that they saw the relevant individual at a specific location and time. In India, where conditions are always changing due to things like light, weather, and specific directions, existing solutions use conventional face acknowledgement computations, which might be problematic because there is no open public contribution. Our research paper employs LBPH, Deep Learning, and Heroku Cloud technologies to construct the system. This application can be used by law enforcement agencies to investigate crime scenes. It is now much quicker, simpler, and more effective to discover the criminal's past behaviour and movements. This makes it easier to comprehend the motives and ideas behind criminal behaviour. Law enforcement agencies get fresh knowledge from these circumstances and are better equipped to act quickly. The opening up of this to the public may facilitate a quicker transfer of information to the agencies. As a result, this approach is more cost- and time-effective. The decision to create the paper as a website rather than an application is driven by the need for greater accessibility and discoverability.Item Encephalon – Solve & Learn(NHCE, 2022) RANJAN KUMAR SAH: SHIVANGI: SHREEYAMCreating a reliable handwritten equation solver using Convolutional Neural Networks (CNN) is a challenging task in image processing and classification. The recognition of handwritten mathematical expressions is one of the most complex challenges in computer vision research because certain characters are segmented and classified, such as polynomial, quadratic, and other mathematical equations. The aim of this study is to develop an application that can solve handwritten equations (arithmetic, quadratic, and trigonometric) and logical operations (logical AND, OR, NOT, NAND, XOR, NOR) using CNN with image processing techniques to achieve high accuracy. Additionally, the application can extract text from images, and users can learn and share educational content on the platform to enhance their knowledge. One of the main advantages of online learning is that it allows users to access updated content and media at any time. The platform also enables users to take tests on various topics and receive instant feedback. Users can code in various languages, including C, Python, and Java. Furthermore, the application provides a way for users to monitor their performance and improve their skills continually. The approach used in this project involves horizontal compact projection analysis and a survey for segmentation and binarization. Connected component analysis and integrated connected component analysis methodologies are utilized for character classification using CNN. Character string operation is used for detecting characters in the polynomial. The experimental results demonstrate the effectiveness of this approach in solving a wide range of mathematical equations. This application aims to make learning an enjoyable experience for students, especially in today's digital age.Item Image De-hazing for Vision Based Applications(NHCE, 2022) NETHRA C: DHAMMINI U: MONIKA M KImages that are captured in haze or fog condition may experience degradation, leading to a loss of accuracy in color, contrast, and visibility. This degradation occurs due to atmospheric particles that scatter and weaken the source radiation. The degree of degradation depends on various factors such as the density of environmental particles, their frequency, and their distance from the capturing device. The current techniques for dehazing images either rely on assumptions to reproduce the transmission map or use a learning system to estimate the dehazed image directly. A recent analysis of popular image dehazing methods using spectral dark images revealed that existing techniques perform poorly with frequency group selection and haze thickness levels. In this study, we propose an effective network called SPIDE-NET that uses spectral and prior-based image dehazing and enhancement techniques. Our approach outperforms existing methods by using spectral dark images from various frequency groups and haze thickness levels. The SPIDE-NET comprises two networks: 1) Spectral Image De-Hazing Network, which is trained on multi-spectral hazy images ranging from 450 nm to 720 nm, and 2) Multiscale Prior-based Image De-hazing Network that uses multi-scale dark channel and variance reduction priors on image trios selected from a multi-spectral hazy image dataset.Item Voice UI(NHCE, 2022) VIDYASAGAR A: YOGESH GOWDA K K: NITHISH S N:Linux Desktop Operating System is the third most popular operating system on the planet.47% of the developer community prefers Linux Desktop OS. Despite its popularity and most of its users being developers, lack a built in Intelligent Virtual Assistant (IVA) in popular distributions like Ubuntu, Mint and Fedora. The proposed framework is an IVA called LARVIS offers an array of functionalities such as file management, fact retrieval, system volume control, internet speed testing, time and date reporting, and more. LARVIS is built using SpeechRecognition, gTTS, and Keras libraries. LARVIS is built so that it can be portable and deployed on any distribution. There are IVAs on other operating systems like Cortana on Windows, Siri on Macintosh, but these are not preferred by most users as it has limited functionality for a desktop Operating System. LARVIS has its own unique ability of navigating through the file system and execute file manipulation tasks, setting it apart from the existing IVAs. LARVIS will improve the user friendliness, efficiency and provide an overall better user experience, which is what determines the success of an operating system, and it also addresses the issue of IVA on this platform.Item Student placement prediction using Machine learning(NHCE, 2022) B YASHWANTH KUMAR; ROHIT G M: BONTHU DEVI VARA PRASAD REDDYAll students dream to obtain a job offer in their hands before they leave their college. A placement chance predictor helps students to have an idea about where they stand and what to be done to obtain a good placement. A placement predictor is a system that could predict the possibility or the type of company a pre-final year student has chances to be placed. Thus a prediction system could help in the academic planning of an institution for future years. With the emergence of data mining and deep learning, many predictor models were introduced by analyzing the previous year student’s dataset. This paper presents a literature survey on different placement prediction models for pre-final year engineering graduate students.Item Student’s Soul - Non Transferable Token(NHCE, 2022) S J KENNITH : TEJASHWIN U : ROHIT MANIVELThe Soul Bound Tokens project is a revolutionary initiative that seeks to immortalize memories, experiences, and legacies in a digital format that is unique, immutable, and decentralized. The project is driven by the need to address the problem of lost intangible assets, which are often lost to time and are not preserved for posterity. These assets represent the essence of a person or community, and their loss means losing a part of our history. The Soul Bound Tokens project leverages the power of blockchain technology to create a decentralized and immutable platform that ensures the immutability and accessibility of these tokens. The platform provides a marketplace for the exchange of these tokens, making it possible for people to own and trade these unique digital assets. The platform will also leverage smart contracts to automate the ownership, transfer, and exchange of these tokens, reducing the transactional costs associated with traditional asset ownership and transfer. The implementation of the project involves several stages, including conceptualization, design, development, testing, launch, and maintenance. The conceptualization phase involves the identification of the key features and functionalities that the platform will offer, while the design phase involves the design of the platform, including the user interface, smart contract architecture, and blockchain infrastructure. The development phase involves the coding of the smart contracts, the implementation of the blockchain infrastructure, and the integration of the user interface. The testing phase involves the testing of the platform to ensure that it is functioning as intended and that it meets the project's objectives. The launch phase involves the deployment of the smart contracts, the integration of the blockchain infrastructure, and the release of the user interface. Finally, the maintenance phase involves the provision of updates, bug fixes, and security patches to ensure the smooth functioning of the platform.Item CYTOCOL – Online Law System(NHCE, 2022) GAGAN P: MAITRI SIDDHARTH MEHTA : NIHARIKA MAHESHThe Indian Penal Code (IPC) is a comprehensive legal document that encompasses a wide range of criminal laws in India. However, navigating the IPC can be a daunting task due to its vast size and complexity. To make legal research more accessible and efficient, the Online Law System has been developed. This web-based platform offers a comprehensive collection of laws organized by keywords and a powerful search function. The system provides legal professionals, researchers, and the public with an invaluable tool for accessing relevant legal information related to criminal activity in India. In this paper, we examine the features and benefits of the Online Law System, highlighting its user-friendly design and extensive collection of legal resources. The Online Law System not only makes legal research more accessible, but also improves the accuracy and efficiency of legal proceedings by providing relevant legal information in a timely manner. This study emphasizes the importance of the Online Law System in facilitating access to legal information and its crucial role in promoting transparency and accountability within the Indian legal system. By providing an extensive collection of laws categorized by keywords and a powerful search function, the Online Law System is a valuable resource for anyone interested in criminal law in India. Overall, the Online Law System is a critical tool for navigating the complexities of the Indian legal system, and its user-friendly design makes it an essential resource for legal professionals, researchers, and the public alike.Item Analysis of Trending NFTs using Time-Series Data and Machine Learning(NHCE, 2022) Kasish S V; Jashwanth M S; Umashankar Reddy MBlockchain technology has reshaped the financial ecosphere. The first recorded use of blockchain may be found in a whitepaper from 2008 that was produced by a person going by the pseudonym Satoshi Nakamoto. NFTs, or Non-Fungible Tokens, are a blockchain product that has sparked a lot of interest from the general public. A NFT is a digital asset that utilizes blockchain technology. Because to its impossibility to be copied, replaced, or divided, it is used to demonstrate ownership and authenticity. The ownership of an NFT is recorded in the blockchain and transferable by the owner, making it possible to buy, sell, and trade NFTs. In this paper, we make an effort to establish a relationship between NFT value and several factors, such as social media, OpenSea data, and others. By using algorithms like Random Forest classifier and Support vector machine we have developed an algorithm which can accurately classify the price range of the NFT taking into consideration the current market trends. Through these adopted algorithms, we have obtained has better accuracy and ability to classify the NFT market. The model has been trained using a large dataset comprising around 70000 records and 130 factors. Thus, the model has shown an accuracy of over 95%, improving the accuracy of previous model already in place. Keywords: Blockchain, NFT, prediction, PCA, SVM, Random Forest classifier, RegressionItem Stroke Risk Prediction With Hybrid Deep Transfer Learning Framework(NHCE, 2022) LAVANYA V ; SANJANA PARIT; SANMITHA KStroke has become a leading cause of death and long-term disability in the world with no effective treatment. Deep learning-based approaches have the potential to outperform existing stroke risk prediction models, but they rely on large well-labeled data. Due to the strict privacy protection policy in health-care systems, stroke data is usually distributed among different hospitals in small pieces. In addition, the positive and negative instances of such data are extremely imbalanced. Transfer learning can solve small data issue by exploiting the knowledge of a correlated domain, especially when multiple source of data are available. In this work, we propose a novel Hybrid Deep Transfer Learning-based Stroke Risk Prediction (HDTL-SRP) scheme to exploit the knowledge structure from multiple correlated sources (i.e., external stroke data, chronic diseases data, such as hypertension and diabetes). The proposed framework has been extensively tested in synthetic and real-world scenarios, and it outperforms the state-of-the-art stroke risk prediction models. Keywords: Stroke Risk Prediction, Deep Transfer Learning, chronic diseases, Deep Learning, Computer Vision.Item Product Price Rate Detection using Regression models and Shopping Assistant(NHCE, 2022) K.V.CHARAN REDDY : KARTHIKEYAN UMESH : G.NIKHILESWARA REDDYE-commerce websites deal with a wide variety of products from all over the world. Our suggested shopping system offers a variety of features to make online buying convenient for users. When a user wants to purchase something from one of these sites, he or she requires instructions on how to utilize the system and other items, much as when shopping in a store. We combine a fake chat system with an e-commerce site in order to deliver these kinds of services online. We also create an app-like interface that allows users to follow and predict product pricing in the future. When a user first visits an online store, he or she can ask questions to learn more about the software. A pattern matching algorithm is used by the e-commerce system to send a customer's query to the AIML Knowledge Base System for an answer. The e-commerce website the user is on will automatically read his product and respond with a response before returning control to the user and the system. This essay focuses on outlining the value of shopping assistants and how they might be improved upon over the current models available.Item Fitness Trainer using Machine Learning and Computer Vision(NHCE, 2022) ADITYA NH : AMAN TEWARI : JYOTHSNA NGood physical fitness has become the need of the hour in an era where it is easy to obtain anything without involving much physical activity. Exercises can be easily found and performed; however, they need to be done with the correct posture to maximize its effectiveness. When not done properly, it can cause injury, pain, be overall ineffective and a waste of time. Posture recognition is a difficult task as it is done on a real time basis. Detection needs to be fast and accurate. Today, many data-sets are available to aid this cause along with a host of machine learning models. This project aims to use a neural network to train a model on a self-created data-set. An attempt has been made to create a web-based application that allows users to learn and practice yoga poses with the help of their internet browser and webcam. A database has been created that is subsequently trained using a neural network. The model is then deployed, and computer vision is utilized to create a skeleton of the user performing the poses. It is then possible to accurately classify the pose. The entire classification process happens within the browser of the user and no additional hardware or software is required. This project further aims to encourage and motivate users towards better fitness by engaging them in simple challenges.Item Plant Infirmity Detection Using VGG-16 Convolutional Neural Network(NHCE, 2022) AAQIB NISSAR : BN SHIVACHARAN : BALAJI SOur project aims to leverage state-of-the-art deep learning techniques to detect plant diseases in agriculture with greater accuracy and timeliness, with a particular emphasis on the Indian economy. Agriculture is a critical aspect of the Indian economy, and plant diseases caused by climatic factors, pests, and weather conditions can significantly impact productivity and yield. Early and precise identification of plant diseases is crucial to estimating agricultural production and maximizing resources while minimizing the use of pesticides. We have used VGG-16 which is a convolutional neural network architecture that uses Rectified Linear Units (ReLU) activation function and have proven to be highly effective and provide high accuracy in detecting and diagnosing plant diseases. Additionally, we have integrated a Graphical User Interface (GUI) that can provide an intuitive platform for users to input plant images and obtain detailed diagnostic information along with an efficient solution. By leveraging these techniques, the research aims to contribute to the early and accurate detection of plant diseases in agriculture, ultimately increasing productivity and yield. This report also discusses potential avenues for further research and the challenges that still exist in identifying plant diseases using available technology.Item Fraud Recruitment Detection Using Machine Learning(NHCE, 2022) AMIDALA VINAY RANGA: AMMIDAL RAKESH: CHARITHA P REDDYIt is getting increasingly difficult to keep safe online due to the daily rise in data and privacy breaches. The quantity of online jobs posted on the internet on various employment portals has significantly increased due to the pandemic crisis. However, some of the internet job postings are actually for bogus positions, resulting in the theft of sensitive data and personal information. The number of people who have fallen victim to bogus job postings is rapidly rising. The bogus employment issue is also getting worse on internet hiring platforms like LinkedIn and Glassdoor. The reputation of the site is hurt by fraudulent job listings. In order to eliminate these bogus jobs, these platforms must detect fraudulent job ads. A machine learning approach is utilized to detect fake posts, which makes use of numerous categorization algorithms. Based on the previously collected data of fraudulent and real job listings, the system would train the model to categorize jobs as true or fake. The difficulty of identifying fraudsters on job advertising can be handled using supervised learning algorithms like classification techniques. We have used two approaches namely machine learning and natural language processing. We use a variety of machine learning algorithms and choose the one that provides the highest accuracy score in determining if a job advertisement headline is real or not. We also can detect whether the job description given by the applicant is fake or not by using natural language processing and observing the accuracy produced on the datasetItem Multi factor crop monitoring and irrigation Optimization(NHCE, 2022) SWAYAM PRAKASH SAHOO: T N CHEHUL CHINNAPPA: VISHVAJITHWith a sizable section of the workforce employed on plantations, the agriculture sector significantly contributes to global employment. India's enormous population relies on agricultural productivity to ensure the safety of its food supply. The nation does, however, confront a number of difficulties, such as a shortage of farmland per capita and poorer production values and land yields in comparison to other wealthy countries. Innovative approaches must be created to boost agricultural productivity while effectively utilizing the available natural resources in order to solve these difficulties. This necessitates a multifaceted strategy that uses cutting-edge farming methods, technology, and the promotion of sustainable practices. India can increase its agricultural output and raise the general standard of living for its people by implementing such measures. Keywords: Cloud computing, IoT, sensors, image processing, irrigation, and precision agricultureItem Phishing Webpage Detection Based on Multidimensional Features Driven by Deep Learning(NHCE, 2022) RAJESH G : RANGANATH: SHREEDHAR V KALAHALPhishing webpage detection is a crucial task in ensuring the security and privacy of online users. Phishing attacks involve fraudulent websites that mimic legitimate ones, with the goal of stealing sensitive information such as usernames, passwords, and credit card numbers. To combat this threat, researchers have developed various techniques for detecting and blocking phishing webpages. In this work, we provide an overview of the state-of-the-art in phishing webpage detection, including both traditional rule-based and machine learning-based approaches. We also discuss the challenges and future directions for improving the effectiveness and efficiency of phishing detection systems. Overall, our analysis highlights the importance of continuing research in this area to better protect online users from the ever-evolving threat of phishing attacks. The detection of phishing webpages is a challenging task, as attackers continuously update their techniques to evade detection. Traditional rule-based approaches are limited in their ability to detect new and evolving attacks, and machine learning-based approaches require large amounts of labeled data to train the models accurately. Recent research has focused on developing more effective phishing detection systems that combine the strengths of both traditional rule-based and machine learning-based approaches. These systems use a variety of techniques, such as analyzing website content and structure, monitoring user behavior, and examining network traffic to detect phishing attacks. Keywords: Phishing, website, labelled data, network traffic, machine learning.