Browse
Recent Submissions
Item Intelligent road quantification system (IRQS)(UMT, Lahore, 2024) Uzair Farid Khan Saddozai; Syed Muhammad Asad Raza Kazmi; Salman Tauheed Bhatti; Muhammad Fasih Tariq; Syed Habeeb Haider ZaidiThe idea for this project is to build an improved road quantification system that tackles the important problem of road surface degradation, primarily concerning the identification and prognosis of potholes and cracks. The goal is to design something that can effectively reach its target of improving the road maintenance task using the assistance of computer vision and machine learning methods. The practical aspect of the study includes the use of imagery from the road surfaces using a dedicated camera that is mounted on a vehicle. The data collected here passes through some preprocessing where the targets, such as pedestrians, vehicles, pavement and others of similar nature are first removed as impurities. The core of the system integrates several key technologies: U-Net is applied in the first step the segmentation, and it is needed to find the geometry and positions of the road. YOLO is then used for detection and gives out the bounding boxes and class probabilities. To increase the accuracy of predictions within successive frames, the use of HMM (Gaussian Mixture Model Hidden Markov Model) is used for the prediction of potholes and cracks. YOLO is re-applied occasionally in order to check and update these predictions so as to maintain perpetuity and efficiency of the detection process. Such, the results indicate that the usage of the described integrated approach enhances the efficiency of the detected road defects and their predictions, making it easier to perform relevant maintenance works as soon as possible. As mentioned earlier, the extracted data is saved in the local storage of the chosen device. Therefore, the workflow of the project allows using the detection, segmentation, and prediction of the road situation to create an effective road monitoring system. This solution deals with one of the most critical and compelling issues of modern society, namely the lack of proper and timely roads maintenance, which improves the safety of roads and their management.Item Addition of variables in DLS by mathematical modeling for better results(UMT, Lahore, 2024) Muhammad Dillawar IkramThe purpose of this research was to investigate the mathematical formulation of the DLS approach, which is used to change the objective score in the sport of cricket. The addition of additional variables that are capable of calculating and integrating the player's ability resources as well as the final outcome of the game was another feature that was included. A comparison will be made between the mathematical model and the DLS approach that is currently in use in order to demonstrate the improvement in the computation of the predicted score.Item Body measurement app (Knitcut)(UMT, Lahore, 2024) Hamid AbbasThis app takes person measurement there dress selection type and some other details its kind of pre data collection providing it to automated system like online shopping, automated tailoring, virtual clothes selection and many more making users friendly environment and making online shopping more efficient in future and applying the edge computing making devise run modl , make calculation independent of cloud processing and providing data integrity and security. Its works by detecting key-points and then mapping high to pixels and then calculate distance between the key points hence taking measurement.Item Safe Eats Gluten Free Guidance(UMT, Lahore, 2024) Ali Haider, Sudais Farooq and Tayyaba AsgharItem AI Surveillance Home Security Robot(UMT, Lahore, 2024) Rohail MansabThe ‘Surveillance Home Security Bot’ project was envisioned to build an improved home security system that is self-operating and intelligent to boost home security by utilizing superior objects’ detection and avoidance. The system has a four-wheel robot with ultrasonic sensors placed at the lateral sides for aiding in the wall-following and obstacle detection. This of hardware arrangement means that the robot is able to move around within a home environment in a manner that is efficient. For the software component of the project, real time object detection is performed using the YOLOv8 model, elements that can be detected include persons, fires and doors. The model is installed on a Raspberry Pi, for its compact form factor and adequate performance to run the YOLOv8 model. Thus, standing for the importance of home security needs and integrating solid hardware with complex software into this project. The application presents a vast enhancement in the supervision and risk identification in a domestic environment at real-time. Some preliminary experiments that were conducted prove that this method is rather accurate and reliable, which makes it possible to consider it suitable for application in practice. Lastly, the project offers a discussion on the findings of a system, the improvement and further development of the final product.Item Language Driven Chatbot for E-Commerce(UMT, Lahore, 2024) Mubashir Irshad, Syed Own Muhammad and Muhammad KamranThe increasing demand for automated customer support in e-commerce has driven the development of intelligent solutions tailored for specific industries. This project presents the design and implementation of an e-commerce chatbot for a gym wear retail website using the Retrieval-Augmented Generation (RAG) model. The chatbot combines retrieval-based and generative approaches to provide accurate, contextually relevant responses to customer inquiries. It utilizes a retrieval system to access product information from a gym wear database and a generative model to craft personalized, coherent replies. The chatbot assists users in navigating the product catalog, answering questions about gym wear clothing, handling order-related queries, and providing recommendations based on customer preferences. The underlying architecture uses GPT-3.5 for generation, ensuring high-quality interactions. The system is evaluated through metrics like response accuracy, user satisfaction, and conversion rates, showcasing its efficiency in improving customer experience in gym wear e-commerce. Our objective is to develop and implement a smart chatbot that enhances the experience of an online user. The chatbot will use NLU and NLG to comprehend and generate text. By utilizing Machine Learning Algorithms, the chatbot would facilitate tailored shopping experience. In addition, the chatbot integrates order tracking that enables convenient tracking, update and real-time notifications. The chatbot also handles customer issues and provides assistanceItem COMPLETE REAL TIME 3D MOTION CAPTURING SYSTEM(UMT, Lahore, 2024) Mansoor Ejaz and Maryam QaisarThis project proposes a novel real-time 3D motion capturing system that leverages the strengths of heterogeneous camera setups and deep learning techniques to achieve high accuracy and high frame rates (fps). Addressing the limitations of existing single-camera systems and computationally expensive multi-camera approaches, our system employs a custom Convolutional Neural Network (CNN) architecture specifically designed for real-time object detection and tracking. By strategically positioning multiple cameras and optimizing data processing pipelines, we aim to overcome challenges like occlusion and varying lighting conditions while maintaining real-time performance. The system Incorporates a comprehensive workflow encompassing data collection, annotation, custom model developing and training, and rigorous evaluation using established metrics. Our approach builds upon the advancements in deep learning and heterogeneous camera systems, offering a significant improvement in real-time 3D motion capturing capabilities for diverse applications.Item ADDITION OF VARIABLES IN DLS BY MATHEMATICAL MODELING FOR BETTER RESULTS(UMT, Lahore, 2024) Muhammad Dillawar IkramThe purpose of this research was to investigate the mathematical formulation of the DLS approach, which is used to change the objective score in the sport of cricket. The addition of additional variables that are capable of calculating and integrating the player's ability resources as well as the final outcome of the game was another feature that was included. A comparison will be made between the mathematical model and the DLS approach that is currently in use in order to demonstrate the improvement in the computation of the predicted score.Item Class Attendance System(UMT, Lahore, 2024) Muhammad Bilal, Salman Shabbir and Maira ArshadThis documentation presents the development, design, working, and implementation of an AI-based attendance system which is sophisticated technology to take attendance in schools, colleges, universities, and offices. In many institutes or organizations RFID biometrics, Punch cards, and traditional sheets are used for attendance but they time time-consuming, have potential inaccuracies, easy to mistake and it’s very difficult to verify attendance by administration. Our proposed model is AI and ML based which takes attendance automatically and is flexible for largescale in which we use Computer Vision, Neural Networks, OpenCV, Convolutional Neural Networks (CNN), Facial Recognition, MTCNN and VGG16. In it, the camera can capture images and match these images with trained images given to the model in the database and store attendance in the database or in an Excel sheet. Our project has a user-friendly interface and ensures data privacy.Item PLANT DISEASE DETECTION SYSTEM(UMT, Lahore, 2024) Shehryar Ahmed and Muhammad Ahmer