NUST Institutions Library Catalogue Search for 'an:119486'

By Afzal, Muhammad . . 134p. , In our contemporary era, information is of prime importance and its dominant use by social media and TV channels for making public opinion and cultural influence is quite evident. Videos form the major portion of media and contain more elaborate information than a single image. Today, videos are piling up in millions every day and their segregation, classification and analysis are upheaval tasks. Live TV video stream contains voice, metadata and image frames full of multiple information including written scripts etc. which can contribute to video classification. But utilization of each type of data we need to do a separate study. However, we have focused on classification of video stream using deep learning (DL) neural networks which are well established solutions for images and small videos classification and gesture recognition. In our study, we have suggested a mechanism for classification of big or live video streams obtained from Pakistani TV News Channels into 5 classes (Advertisement, News, Talk Show, Sports & Entertainment Program) using supervised DL pretrained neural networks. Due to non-availability of authentic dataset on this subject, we have created a customized data of videos recorded (approximately 335 hours videos) from various sources like different TV channels‘ websites and YouTube. Videos were processed to extract image frames to prepare a trainable dataset. For our experimentation, we have mainly used pretrained ResNet variants (ResNet18, ResNet34, ResNet50, ResNet101 & ResNet152) on ImageNet dataset and few other models like AlexNet, ConvNeXt_Tiny, DenseNet121, SqueezeNet and VGG11 for comparison purposes. Then modified the last classification layer of the network as per number of target classes and finetuned all weights of neural network on the subject dataset. We carried out various experiments on these neural networks and achieved quite encouraging results having accuracies ranging from 95% to 99%. For testing of videos on trained models, dynamic averaging time domain window was applied to diminish the jitters in the output results. This can be useful in many other applications as well including social media & advertisements analysis, classification of small videos, industrial and business automation etc. 30cm.

Place Hold on Classification of Live Video Stream from Pakistani News Channels (Urdu) using Deep Learning Latest Techniques /

Focus and Engagement Level Detection Using Computer Vision and Machine Learning in a Classroom Environment /

By Poonja, Hasnain Ali . . 64p. , Due to Covid 19, the global education system has changed toward online learning, which has a high dropout rate. Therefore, it is vital that students maintain their level of interest. Therefore, detection of engagement level alone is insufficient for analyzing and improving learning and teaching techniques. To promote student engagement in STEM and online learning environments, technologies such as AR/VR and Haptics should be implemented. Utilizing facial emotion, body pose, and head rotation, a web-based computer vision system is developed and implemented to identify student involvement levels using webcams during tasks such as online classrooms, haptic interaction, and augmented reality. In addition, an AR and Haptics-based World Map is being designed and developed. To evaluate and compare three types of learning scenarios, namely (1) Traditional, (2) Augmented Reality-based, and (3) Haptics-based, two methods are employed: (1) Trained Computer Vision models are tested for 3 scenarios, and (2) A user study is conducted using the Positive and Negative Affect Schedule (PANAS) Questionnaire and NASA-Task Load Index, from which conclusions are drawn. The results of a comparison of Traditional, Augmented reality, and Haptics-based learning indicate that Haptics and Augmented Reality-based learning are the most immersive and increase levels of engagement during online learning and STEM training, whereas Traditional learning methods are the least effective during online classes. User studies and computer vision models are utilized to validate the results. 30cm.

Place Hold on Focus and Engagement Level Detection Using Computer Vision and Machine Learning in a Classroom Environment /

Lung damage estimation through ground glass opacity detection from 3D reconstructed HRCT scans /

By Naeem, Abdul Hanan . . 55p. , High-resolution computed tomography (HRCT) scans have become an essential tool for the diagnosis of lung diseases, especially during the COVID-19 pandemic. However, the manual analysis of these scans by clinicians can be time-consuming and error-prone, leading to delayed diagnosis and treatment. In this thesis, we present a deep learning-based system for the automated estimation of lung damage through the detection of ground-glass opacities (GGOs) using 3D reconstructed HRCT scans. The system utilizes a MobileNetV3 backbone combined with a Lite Reduced Atrous Spatial Pyramid Pooling (LR-ASPP) segmentation head to accurately segment GGO regions in the lung. The 3D reconstruction of the scans helps to provide clinicians with a more comprehensive view of the lungs, allowing for better identification and analysis of GGOs. To train and evaluate our system, we utilized a custom dataset of HRCT scans. The results demonstrate that our system achieved high accuracy in detecting and segmenting GGO regions in the lungs, with an overall IOU of 0.62. Additionally, our system was able to provide clinicians with a more efficient method for analyzing HRCT scans, reducing the time required for diagnosis and allowing for earlier detection of lung diseases. In conclusion, our deep learning-based system provides a promising approach for the automated estimation of lung damage through GGO detection using 3D reconstructed HRCT scans. By combining state-of-the-art techniques in deep learning and medical imaging, our system can provide clinicians with an accurate and efficient method for analyzing HRCT scans, potentially leading to improved patient outcomes and reducing the burden on healthcare systems. 30cm.

Place Hold on Lung damage estimation through ground glass opacity detection from 3D reconstructed HRCT scans /

Pose-Based Seamless Video Stitching for Real World Applications /

By Hassan, Salman . . 70p. , Combining videos of humans performing different gestures in a smooth way can potentially have many uses across a wide range of fields. These include entertainment, virtual reality, robotics, education, & communication. The goal of this research work is set in this context. This research focuses on developing a system that takes individual videos of humans performing motion gestures, and stitches them in a way that minimizes spatial discontinuities between upper torso joints, thus joining two or more human gestures into one seamless continuous motion. It begins by investigating & comparing current frameworks used to stitch individual human motion gestures and investigates the theoretical and mathematical approaches behind them, proceeding in a step-by-step way. First, it collects sign videos for most commonly used English sentences of lengths 2-8. Then, it preprocesses these videos to convert them into a standardized form. Following that, it extracts landmarks to prune unnecessary parts of videos. It then calculates human joint coordinates using pose estimation. After that it calculates link vectors and human shoulder, and elbow angles using linear algebra. Following that, the system interpolates joint coordinates at transitions between signs and uses them to calculate interpolated joint angles. Concurrently, actual joint coordinates are used to calculate actual joint angles, which are then used to calculate wrist poses using forward kinematics. These wrist poses are compared with the same obtained from feeding interpolated joint angles to forward kinematic models. An ablation study was then conducted that measured mean errors across different combinations of spline degree, percentage of knots, & length of sentences. LSQ Univariate Spline with degree 4, knots percentage of 90%, and sentence length of 4 produced least mean error. Transition errors (errors between sign transitions were also calculated & recorded for each of 100 sentences. In this way, smoothness of different interpolating functions was quantified 30cm.

Place Hold on Pose-Based Seamless Video Stitching for Real World Applications /

Stitch Multiple Images for Generating Quality Panorama /

By Riaz, Sibgha . . 53p. , Stitching multiple images for achieving the 360 view of any environment is a challenging task. Traditionally, the whole process of image stitching is based on distinctive features that are very helpful for estimating the other parameters of the whole algorithm. As different images require different suitable parameters or weights for achieving the best results and we need to predict those suitable parameters for each case independently. In our proposed model first small neural network based techniques are implemented that are just used for estimating the quality panorama hyper parameters and then we apply the whole stitching algorithm on sample images by using those predicted parameters. Therefore, due to lack of labeled data we are unable to train any supervised model for those hyper parameter selection that’s why we build an unsupervised technique that makes decisions based on just extracted features quality, confidence and count of inliers etc. By estimating the good parameters we are able to stitch a quality panorama that doesn't have any ghosting artifacts, blending discontinuities, seamless and alignment errors as well. We evaluate the performance of our proposed model on three datasets and analyze performance in both perspective quality and computational time and conclude that our model outperforms with other state of the art stitching algorithms in both perspectives. 30cm.

Place Hold on Stitch Multiple Images for Generating Quality Panorama /

Ball Interception in Humanoid Robot Soccer /

By Khan, Saman . . 70p. , Humanoid Robots are widely used in various real-world applications which contributed in raising the interest of researchers to enhance Robot development and deployment in dynamic environments. Robocup Soccer league is a platform for qualitative development in the fields of Robot Kinematics and Dynamics, Motion Planning and Control, Navigation, Computer Vision and Machine Learning. New challenges are introduced each year for teams that aim to improve robot behaviors in real world scenarios. One of the most popular leagues in Robocup is Standard Platform League that majorly focuses on improving individual robot capabilities, team capabilities and coordinated strategies in dynamic soccer matches played by teams of Aldebaran Nao humanoid robots. The robots with capability to quickly gain ball possession from opponent by intercepting the moving ball path will contribute to a better defense strategy. The aim of this study is to improve individual robot and team skills during game scenarios where ball is moving using motion primitives and geometric parameters. The proposed strategy is used to compute the interception point using geometric equations given the ball position and velocity information. The robot response is dependent on factors including initial position and velocities of robot and ball. Furthermore, to demonstrate the feasibility of proposed approach, different scenarios with single and multi-agents are simulated using SimRobot. Simulations validate the effectiveness of proposed algorithm in moving ball interception scenario. 30cm.

Place Hold on Ball Interception in Humanoid Robot Soccer /

An Unreal Engine Based Human Robot Interaction Framework /

By Asif, Muhammad Hassaan . . 64p. , Existing frameworks lack the support for testing Human Robot Interaction (HRI) research which in turn often has to be tested practically making it time consuming and expensive. To overcome this issue, a HRI framework based on Unreal Engine is proposed which consists of a virtual Nao or Pepper robot along with virtual humans with verbal and non-verbal behaviours in an environment. Machine Learning (ML) algorithms along with the behaviour of the virtual robot in response to the interaction with the virtual humans and the environment can be programmed using a Python API which communicates with Unreal Engine C++ in real time. Several experiments related to multiple aspects of HRI: (1) Verbal Interaction; (2) Non-Verbal Interaction; (3) Emotional Interaction were conducted in both the virtual and real world environments and the results were compared to validate the feasibility of the framework. A Reinforcement Learning (RL) algorithm was also tested to further indicate the usefulness of the framework. Through the use of a Virtual Reality (VR) headset, a human can be immersed in the framework to interact with the robot in real time 30cm,.

Place Hold on An Unreal Engine Based Human Robot Interaction Framework /

Perception of Emotion in Human-Robot Interaction /

By Zia, Muhammad Faisal. . 59p. , Perception of emotion is an intuitive replication of a person’s internal state without the need for verbal communication. Visual emotion recognition has been broadly studied and several end-toend deep neural networks (DNNs)-based and Machine learning-based models have been proposed but they lack the ability to be implemented in low-specification devices like robots, and vehicles. The drawbacks of conventional handcrafted feature-based Facial Emotion Recognition (FER) methods are eliminated by DNNs-based FER approaches. In spite of that, Deep Neural Network based FER techniques suffer from high processing costs and exorbitant memory requirements, their application is constrained in fields like Human-Robot Interaction (HRI) and HumanComputer Interaction (HCI) and relies on hardware requirements. In aforementioned study, we presented a computationally inexpensive and robust FER system for the perception of six basic emotions (i.e., disgust, surprise, fear, anger, happy, and sad) that is capable of running on embedded devices with constrained specifications. In the first step after pre-processing input images, geometric features are extracted from detected facial landmarks, considering the facial spatial position among influential landmarks. The extracted features are given as input to trainthe SVM classifier. Our proposed FER system was trained and evaluated experimentally using two databases, Karolinska Directed Emotional Faces (KDEF) and Extended Cohn-Kanade (CK+) database. Fusion of KDEF and CK+ datasets at the training level were also employed in order to generalize the FER system’s response to the variations of ethnicity, race, national and provincial backgrounds. The results show that our proposed FER system is optimized for real-time embedded applications with constrained specifications and yields an accuracy of 96.8%, 86.7% and 86.4% for CK+, KDEF and fusion of CK+ and KDEF databases respectively. As a part of our future research objectives, the developed system will make a robotic agent capable of perceiving emotion and interacting naturally without the need for additional hardware during HRI. 30cm.

Place Hold on Perception of Emotion in Human-Robot Interaction /

Glossing of American Sign Language (ASL) /

By Thaheem, Muhammad Arslan . . 49p. , American Sign Language (ASL) is a sign language used in America with slight modifications across different regions. Over the years it has developed and included a lot of new signs in it. In order for the deaf community to take notes and communicate with common people, ASL glossing is done which is an organized sentence structure of ASL. The goal of this research is to make a rule-based engine that can convert English sentences into ASL Gloss. The research included three phases. Firstly, we collected the English to ASL sentences from different resources including books, websites, etc. Secondly, we made rule-based engine to standardize the format of glossing. Since there are no proper set of rules written till now for ASL, we extracted common rules from sentences collected from different sources and continued our research on the basis of that. Thirdly, the engine was checked using different English sentences. We were able to achieve a BLEU score of 20.85 on the test set. 30cm.

Place Hold on Glossing of American Sign Language (ASL) /

Realtime Automatic Vehicle License Plate Recognition /

By Mansoor ,Ahmad. . 54p. ; 33cm..

Place Hold on Realtime Automatic Vehicle License Plate Recognition /

Flight Stability of UAV-Manipulator System for Grasping Applications using Sliding Mode Control /

By Sial, Munadi Ahmad . . 122p. , An Aerial Manipulator is a coupled system which consists of a robotic arm attached to the body of a drone. A drone is itself a non-linear system which is linearly approximated to enable PID control. With the addition of a manipulator, the non-linearity is increased to a significant extent. The combined system has very complex dynamics due to the shifting of the center of mass from actuating the arm. This places a great demand on the flight control system in order to maintain stability. Another problem is the sudden change in system dynamics whenever an object is grasped. The benefit of such a complex system is that the drone will be able to physically interact with the environment that it is inspecting. The interaction can be in the form grasping, pushing, pulling and transport of objects. The purpose of this research is to design and control an aerial manipulator for the stable grasping of small objects. 30cm.

Place Hold on Flight Stability of UAV-Manipulator System for Grasping Applications using Sliding Mode Control /

Data Association Network /

By Khizer ,Hood . . 50p. ; 33cm..

Place Hold on Data Association Network /

Instance Segmentation of Laparoscopic Instruments and Organs /

By RIAZ ,AQSA. . 47p. ;

Place Hold on Instance Segmentation of Laparoscopic Instruments and Organs /

Solving Inverse Kinematics Using Artificial Neural Networks /

By Mohsin, Maaz. . 46p. ; 33cm..

Place Hold on Solving Inverse Kinematics Using Artificial Neural Networks /

Stress Monitoring Using Wearable Sensors /

By TAHIR, ANISA . . 94p. ; 33cm..

Place Hold on Stress Monitoring Using Wearable Sensors /

HAPTIC FEEDBACK FOR REHABILITATION SYSTEMS /

By Yunus ,Rumshaa . . 60p. ; 30cm..

Place Hold on HAPTIC FEEDBACK FOR REHABILITATION SYSTEMS /

Overhead Ball Throwing for Humanoid Robot Soccer /

By MASOOD ,QAZI AAFAQUE . . 42p. ; 33cm..

Place Hold on Overhead Ball Throwing for Humanoid Robot Soccer /

Development of a Generalized Whole Body Motion Architecture for a Humanoid Robot with Applications in Soccer Robotics /

By Saifullah. . 94p. ; 33cm..

Place Hold on Development of a Generalized Whole Body Motion Architecture for a Humanoid Robot with Applications in Soccer Robotics /

EYE MOVEMENT BEHAVIOUR VARIES WITH EMOTIONAL CUES /

By KHALID, USMAN . . 69p. ; 33cm..

Place Hold on EYE MOVEMENT BEHAVIOUR VARIES WITH EMOTIONAL CUES /

Bi-directional Robotic Communicator for Hearing Impaired /

By Fayyaz, Sobia .

Place Hold on Bi-directional Robotic Communicator for Hearing Impaired /

BALL DETECTION ROBOT DETECTION SELF LOCALIZATION AND PROBABILISTIC TRACKING OF A ROBOTS IN ROBOCUP STANDARD PLATFORM LEAGUE /

By ASHFAQ, ASBAH . . 41p. ; 33cm.

Place Hold on BALL DETECTION ROBOT DETECTION SELF LOCALIZATION AND PROBABILISTIC TRACKING OF A ROBOTS IN ROBOCUP STANDARD PLATFORM LEAGUE /

VISION BASED FINGER INTERACTION WITH PROJECTOR CAMERA CALIBRATION /

By WAJHI ,USAMA . . 82p. ; 33cm..

Place Hold on VISION BASED FINGER INTERACTION WITH PROJECTOR CAMERA CALIBRATION /

Automatic Detection and Recognition of Citrus Fruits Diseases Using Deep Learning Model /

By Faisal, Shah . . 59p. , In a country's economy, agriculture plays a very vital role. Agriculture's yield and production are reduced by plant diseases, resulting in significant economic losses and instability in the food market. In plants, the citrus fruit crop is one of the most important agricultural products in the world, produced and grown in around 140 countries. It has a lot of nutrients, such as vitamin C. However, due to pests and diseases, citrus cultivation is widely affected and suffers significant losses in both yield and quality. The majority of plant diseases exhibit visible symptoms, and the accepted method used today is for a skilled plant pathologist to detect the diseases by examining affected plant leaves under a microscope, which is a costly and time-consuming method. During the last decade, computer vision and machine learning have been widely adopted to detect and classify plant diseases, providing opportunities for early disease detection and bringing improvements to agricultural production. The early detection and accurate diagnosis of plant diseases are essential for reducing their spread and damage to crops. In this work, we presented an automatic system for early detection and recognition of citrus plant diseases based on a deep learning (DL) model to improve accuracy and reduce computational complexity. The most recent transfer learning-based models were applied to our dataset in order to increase classification accuracy. In this work, we successfully proposed a CNN-based pre-trained model (EfficientNetB3, ResNet50, MobiNetV2, (InceptionV3) for the identification and classification of citrus plant diseases using transfer learning. In order to assess the performance of the model, we found that the transfer of an EfficientNetb3 model led to the highest training, validating, and testing accuracies, which were 99.43%, 99.48%, and 99.58%, respectively. The proposed CNN model exceeds other cutting-edge CNN network architectures developed in earlier literature in the identification and categorization of citrus plant diseases. 30cm.

Place Hold on Automatic Detection and Recognition of Citrus Fruits Diseases Using Deep Learning Model /