Muhammad Imad
@ulster.ac.uk
Computing
Ulster University UK
RESEARCH INTERESTS
Image Processing, Machine Learning, Medical Imaging, Deep Learning, Computer Vision, Natural Language Processing
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Haris Anjum, Usama Arshad, Raja Hashim Ali, Zain Ul Abideen, Muhammad Huzaifa Shah, Talha Ali Khan, Ali Zeeshan Ijaz, Abu Bakar Siddique, and Muhammad Imad
IEEE
Facial Liveness Detection is instrumental in combating fraudulent practices and identity theft by differentiating genuine faces from forgeries. Given that facial recognition is now an integral part of many sectors like banking and law enforcement, liveness detection has become a vital aspect for maintaining the trustworthiness of these applications. Unfortunately, the utility of prevalent facial liveness detection models is constrained by their lack of generalizability to novel, unseen data, which presents a substantial impediment to practical applications. To mitigate this issue, our research endeavored to construct a reliable and precise facial liveness detection model that can effectively navigate this challenge and ensure consistent performance across diverse datasets. The devised model exhibited excellent accuracy, which attests to its robustness. This research underscores the advantages of employing deep learning methods such as Convolutional Neural Networks (CNNs) as they significantly enhance the model’s accuracy. We intend for our findings to drive further research and progress in this critical field, by laying the groundwork for future work that integrates multiple liveness detection methods and collects more varied training data to develop an even more resilient and efficient system.
Mahnoor Iftikhar, Raja Hashim Ali, Memoona Saleem, Nazia Shahzadi, Usama Arshad, Talha Ali Khan, Ali Zeeshan Ijaz, Nisar Ali, and Muhammad Imad
IEEE
In the vast expanse of global travel, the intricate web of interconnected airline networks serves as the lifeline of modern aviation. Amidst the bustling realm of global aviation, where time and resources are of the essence, the optimization of airline networks emerges as a paramount pursuit, unleashing the power to streamline operations, enhance connectivity, and elevate the travel experience for passengers worldwide. In this paper, four prominent graph-based algorithms, namely A*, Bellman-Ford, bidirectional search, and Dijkstra’s algorithm, were used and compared for optimization of airline networks. These algorithms leverage the inherent structure of the airways network, represented as a graph, to determine the most efficient routes. By evaluating the execution times, path lengths, and complexities of these algorithms, this research equips the airline industry with valuable insights for optimizing airways path-finding. With optimized path-finding, the skies become more navigable, fostering a seamless travel experience for passengers and bolstering the economic growth driven by the aviation sector.
Raja Hashim Ali, Ali Zeeshan Ijaz, Muhammad Huzaifa Shah, Nisar Ali, Muhammad Imad, Said Nabi, Kiran Perveen, Javaria Tahir, and Memoona Saleem
IEEE
This paper explores the integration of Virtual Reality and Cloud Computing to develop an innovative app for interior design. Virtual Reality represents a groundbreaking advancement in the digital realm, offering a transition from 2D to 3D environments that closely resemble real-life simulations. On the other hand, Cloud Computing revolutionizes traditional approaches to server management, storage, networking, and software by providing these services over the Internet. Our objective was to leverage the capabilities of both technologies to create an interior design app that allows users to conveniently renovate and modify architectural spaces from anywhere in the world, regardless of their physical location.
Mahnoor Iftikhar, Raja Hashim Ali, Memoona Saleem, Usama Arshad, Ali Zeeshan Ijaz, Nisar Ali, Muhammad Imad, Muhammad Abu Bakar, and Ali Aftab
IEEE
The exponential growth in the adoption of information and communication technologies has sparked a notable surge in the demand for Natural Language Processing (NLP) tools. The tagging/identification of Part-of-Speech (POS) is of utmost importance in numerous natural language processing applications, including information extraction, parsing, and machine translation. It entails the assignment of a distinct part of speech to every word within a given corpus. The Hidden Markov Model (HMM) stands as the prevailing technique employed for Part-of-Speech tagging. The foundation of this approach lies in a probabilistic model that effectively captures the intricate relationships between words and their corresponding tags in a sequential manner. In this research, the Hidden Markov Model (HMM) is investigated for its potential application in POS tagging. It showcases the dependability and efficacy of this approach by attaining an impressive accuracy rate of 96.67%. Also, the precision, recall, and F-score metrics achieved stand at 96%, 96.5%, and 95% respectively, further solidifying the effectiveness of the HMM method.
Attia Shabbir, Raja Hashim Ali, Muhammad Zeeshan Shabbir, Zain Ul Abideen, Talha Ali Khan, Ali Zeeshan Ijaz, Nisar Ali, Muhammad Imad, and Muhammad Abu Bakar
IEEE
Breast cancer is a significant global healthcare challenge, particularly in developing and underdeveloped countries, with profound physical, emotional, and psychological consequences, including mortality. Timely diagnosis and accurate treatment are crucial in addressing this issue. We propose the utilization of a feature selection technique to identify the most relevant features from among all features for breast cancer diagnosis, and show that Genetic Algorithms are impressive for this task. The study compares the results of GA with no selection and an alternative method, Principle Component Analysis (PCA). Three machine learning models, all based on supervised learning with data split into training and test data, are employed for binary classification using the selected feature subset. The evaluation metrics employed encompass accuracy, precision, recall, and F1-score. Among the selected models, Random Forest demonstrates the most favorable outcomes, achieving an accuracy score of 0.96, precision score of 0.96, recall value of 0.98, and an F1-score of 0.97. These results underscore the effectiveness of GA in feature selection for breast cancer diagnosis. Consequently, the integration of Genetic Algorithms (GA) with Random Forest showcases the superior performance among the evaluated models.
Omar Kashif Majeed, Raja Hashim Ali, Ali Zeeshan Ijaz, Nisar Ali, Usama Arshad, Muhammad Imad, Said Nabi, Javaria Tahir, and Memoona Saleem
IEEE
This paper presents an analysis of solving the N-Queen problem using a genetic algorithm and compares its performance with traditional search algorithms like breadth-first search (BFS) and depth-first search (DFS). The N-Queen problem is a famous problem in the field of artificial intelligence that has been studied in depth. The genetic algorithm is an optimization algorithm inspired by the process of natural selection and evolution in living organisms. BFS and DFS are classical search algorithms that explore the search space to find a solution. This paper discusses all the main components of the genetic algorithm, such as population generation, fitness function, selection, crossover, and mutation, and also explains how it is used to solve the N-Queen problem. Traditional search methods, i.e., BFS and DFS, are also briefly discussed. The experimental results show that classical searching approaches are better for small-sized problems, but the genetic algorithm outperforms BFS and DFS in computation time for medium and large-sized problems. This paper concludes by analyzing the effect of parameter tuning on the genetic algorithm and suggests possible future work in the field.
Abu Bakar Siddique, Muhammad Abu Bakar, Raja Hashim Ali, Usama Arshad, Nisar Ali, Zain Ul Abideen, Talha Ali Khan, Ali Zeeshan Ijaz, and Muhammad Imad
IEEE
Feature selection is a critical factor affecting the performance of optimization algorithms. Without proper feature selection, optimization algorithms may suffer from slow convergence, overfitting, increased computational requirements, and longer execution times. On the other hand, omitting important features can lead to loss of relevant information, decreased accuracy, bias, and increased vulnerability to noise and outliers. This study investigates the use of genetic algorithms as a feature selection technique for a classification problem, specifically the mushrooms classification problem. Random forest is employed as the machine learning classifier, and genetic algorithms are compared with correlation as the feature selection method. The results show that genetic algorithms achieve higher accuracy, precision, recall, and F1-score compared to correlation-based feature selection. However, genetic algorithms have limitations in their applicability to specific optimization problems, the need for proper parameter setup, and longer convergence times. Despite these drawbacks, genetic algorithms prove to be superior to other feature selection techniques, particularly correlation-based approaches. This study highlights the importance of selecting appropriate feature selection techniques for optimization algorithms to improve their performance and achieve better results. In addition, this study explored the performance of various machine learning approaches on the complete mushroom dataset with 22 features and shows that genetic algorithms with feature selection as the most accurate method.
Muhammad Huzaifa Shah, Muhammad Abu Bakar, Raja Hashim Ali, Zain Ul Abideen, Usama Arshad, Ali Zeeshan Ijaz, Nisar Ali, Muhammad Imad, and Said Nabi
IEEE
This study investigates the application of the Mutual Information (MI) feature selection technique to improve the accuracy of Machine Learning (ML) models on NSL-KDD datasets, building upon prior research. Six ML models, namely Decision Tree (DT), Logistic Regression (LR), K-Nearest Neighbor (KNN), Random Forest (RF), Naive Bayes (NB), and Support Vector Machine (SVM) with different kernels (1st, 2nd, and 3rd), are implemented for classification purposes. The proposed DT model in this study shows higher accuracy than the DT model proposed in the original paper by Ingre et al. for Intrusion Detection System (IDS). Additionally, a multi-class classification model for NSL-KDD datasets is developed, considering both normalized and non-normalized features. Interestingly, it is observed that the models trained without normalized features achieve higher accuracies compared to those trained with normalized features. Moreover, the study enhances the classification performance of the DT-based IDS using the Correlation based Feature Selection (CFS) technique for feature selection. The proposed IDS is evaluated both before and after feature selection for multi-class classification (normal and various attack types) and binary classification (normal and abnormal data).
Fatima Faridoon, Raja Hashim Ali, Zain Ul Abideen, Nazia Shahzadi, Ali Zeeshan Ijaz, Usama Arshad, Nisar Ali, Muhammad Imad, and Said Nabi
IEEE
PolyCystic Ovary Syndrome (PCOS) is a hormonal disorder frequently found in women of reproductive age having a significant impact on the cause of infertility. It is an endocrine condition characterized by abnormalities in female hormone levels and aberrant synthesis of male hormones. This syndrome causes ovarian malfunction, increasing the risk of miscarriage and infertility. PCOS has a wide range of symptoms, making a diagnosis difficult. In this study, we proposed a feature selection method based on genetic algorithm with logistic regression to increase the accuracy of early diagnosis. A dataset containing the clinical and biochemical characteristics of 109 patients, including 36 with PCOS, is used. The genetic algorithm is used to extract the most important features from the dataset. Then, a logistic regression model is used for classification. The proposed model outperformed the baseline model that used all features, which had an accuracy of 86.2%, and produced a markedly improved accuracy of 95.4%. These results show that the proposed model effectively locates key characteristics for PCOS diagnosis.
Adnan Haider, Abu Bakar Siddique, Raja Hashim Ali, Muhammad Imad, Ali Zeeshan Ijaz, Usama Arshad, Nisar Ali, Memoona Saleem, and Nazia Shahzadi
IEEE
Social media platforms serve as the main conduit for information and communication in today’s society. Social media platforms have ingrained themselves into our daily lives, and as access is expanded to more distant regions, their user bases are rapidly growing. Roman Urdu is the primary form of communication on social media among the 71.70 million users in Pakistan. As a result of these advancements and the rise in users, cyberbullying—also known as digital bullying—has increased. This study emphasizes on social media users who interact using Roman Urdu, an Urdu dialect written using the English alphabet. In this study, we investigated the issue of online bullying behavior on the data (Roman Urdu) that was collected from Kaggle. This is one of the rare research projects that, as far as we are aware, addresses Roman Urdu cyberbullying behavior. The purpose of the research we propose is to find a suitable model for Roman Urdu cyberbullying behavior classification. The data was in raw form in a way that the contents and data annotations were in the different lists inside different dictionaries of the root dictionary. After labeling the data, the dataset has then undergone pre-processing to remove duplications, stop words, punctuation, and other sources of noise. Then, based on supervised learning, a set of different learning algorithms, including Support Vector Machine, Decision Tree, Random Forest, Naïve Bayes, and Logistic Regression were applied to both types of extracted features. The features are extracted using two separate approaches, Count-Vectorizer and TF-IDF (term frequency-inverse document frequency) Vectorizer. With performance rates of 89.9% when applied to TF-IDF features and 88.3 percent when applied to CV features, Random Forest (RF) outperformed the other developed algorithms by both combinations. The suggested approach might assist online social apps and chat rooms create stronger bully word filters and make the internet a safer place for users.
Naimullah Naim, Muhammad Imad, Muhammad Abul Hassan, Muhammad Bilal Afzal, Shabir Khan, and Amir Ullah Khan
European Alliance for Innovation n.o.
From the last decades different types of network schemes are pitched to enhance the user performance. Software Defined Networks (SDN) is also considered as important factor for different network schemes and its proper administration or management. Due to major deployment in today’s networking era SDN are further sub divided in to commercial and open-source controllers. Commercial and open-source controllers are utilized in different type of businesses. According to our knowledge considerable amount of literature is available on these controllers but did not provide or analyse performance of these controllers on different network parameters. This paper evaluates and compares the performance of two well-known SDN open-source controllers POX and RYU with two performance assessments. The first assessment is the implementation of optimal path by using Dijkstra's algorithm from source to destination. Second assessment is the creation of a custom topology in our desired tool (MiniNet emulator). Then, the performance in terms of QoS parameters such as Jitter, throughput, packet loss, and packet delivery ratio are computed by two end hosts in each network. After the assessments, the performance of POX are optimal as compare to the RYU and best suited to be deployed in any scenario.
Muhammad Imad, Zahoor Ali Khan, Shah Hussain Bangash, Irfan Ullah Khan, Sheeraz Ahmad, and Atif Ishtiaq
IEEE
Skin cancer is one of the most common and dangerous diseases due to a lack of awareness of its signs and methods for prevention. Skin cancer disease can be counted as a fourth burden disease around the world, with the rate of deaths dramatically growing globally. Therefore, early detection at an early stage is necessary to stop the spread of cancer. In this paper, we detect and classify multi-label skin cancer and implement the optimal techniques using machine learning and image processing approaches. However, preprocessing methods assist in removing irrelevant and unnecessary features from the label encoder, and standard features are applied to standardize the range of functionality by scaling the input variance unit. Moreover, various machine learning techniques were applied to check the performance of every classifier on the HAM10000_metadata dataset. The experimental analysis was conducted on the HAM10000_metadata dataset, which consists of seven different types of skin cancer. The results analysis shows that machine learning algorithms such as SVM, DT, and GNB obtained the highest accuracy compared to the other classifiers.
Shah Hussain Badshah, Muhammad Imad, Irfan Ullah Khan, and Muhammad Abul Hassan
Springer International Publishing
Muhammad Imad, Oualid Doukhi, Deok Jin Lee, Ji chul Kim, and Yeong Jae Kim
MDPI AG
Feasible local motion planning for autonomous mobile robots in dynamic environments requires predicting how the scene evolves. Conventional navigation stakes rely on a local map to represent how a dynamic scene changes over time. However, these navigation stakes depend highly on the accuracy of the environmental map and the number of obstacles. This study uses semantic segmentation-based drivable area estimation as an alternative representation to assist with local motion planning. Notably, a realistic 3D simulator based on an Unreal Engine was created to generate a synthetic dataset under different weather conditions. A transfer learning technique was used to train the encoder-decoder model to segment free space from the occupied sidewalk environment. The local planner uses a nonlinear model predictive control (NMPC) scheme that inputs the estimated drivable space, the state of the robot, and a global plan to produce safe velocity commands that minimize the tracking cost and actuator effort while avoiding collisions with dynamic and static obstacles. The proposed approach achieves zero-shot transfer from a simulation to real-world environments that have never been experienced during training. Several intensive experiments were conducted and compared with the dynamic window approach (DWA) to demonstrate the effectiveness of our system in dynamic sidewalk environments.
Muhammad Abul Hassan, Sher Ali, Muhammad Imad, and Shaista Bibi
Springer International Publishing
Muhammad Imad, Muhammad Abul Hassan, Shah Hussain Bangash, and Naimullah
Springer International Publishing
Muhammad Abul Hassan, Muhammad Imad, Tayyabah Hassan, Farhat Ullah, and Shaheen Ahmad
Springer International Publishing
Adnan Hussain, Muhammad Imad, Asma Khan, and Burhan Ullah
Springer International Publishing
Muhammad Imad, Adnan Hussain, Muhammad Abul Hassan, Zainab Butt, and Najm Ul Sahar
Springer International Publishing
Muhammad Imad, Oualid Doukhi, and Deok-Jin Lee
MDPI AG
Three-dimensional object detection utilizing LiDAR point cloud data is an indispensable part of autonomous driving perception systems. Point cloud-based 3D object detection has been a better replacement for higher accuracy than cameras during nighttime. However, most LiDAR-based 3D object methods work in a supervised manner, which means their state-of-the-art performance relies heavily on a large-scale and well-labeled dataset, while these annotated datasets could be expensive to obtain and only accessible in the limited scenario. Transfer learning is a promising approach to reduce the large-scale training datasets requirement, but existing transfer learning object detectors are primarily for 2D object detection rather than 3D. In this work, we utilize the 3D point cloud data more effectively by representing the birds-eye-view (BEV) scene and propose a transfer learning based point cloud semantic segmentation for 3D object detection. The proposed model minimizes the need for large-scale training datasets and consequently reduces the training time. First, a preprocessing stage filters the raw point cloud data to a BEV map within a specific field of view. Second, the transfer learning stage uses knowledge from the previously learned classification task (with more data for training) and generalizes the semantic segmentation-based 2D object detection task. Finally, 2D detection results from the BEV image have been back-projected into 3D in the postprocessing stage. We verify results on two datasets: the KITTI 3D object detection dataset and the Ouster LiDAR-64 dataset, thus demonstrating that the proposed method is highly competitive in terms of mean average precision (mAP up to 70%) while still running at more than 30 frames per second (FPS).
Nazish, Syed Irfan Ullah, Abdus Salam, Wajid Ullah, and Muhammad Imad
Springer International Publishing
COVID-19 is an infectious disease caused by SARS-Cov2 that has spread rapidly worldwide. According to the World Health Organization (WHO), the total cases of 4374783839 are reported from different countries. In this consequence, it is necessary to diagnose automatically COVID-19, which helps in prevention during spreading among people. In this study, we have used machine learning techniques to diagnose and classify the COVID-19 and normal patients from chest X-ray images using a machine learning technique. The proposed system involves pre-processing, feature extraction, and classification. In the pre-processing, the image is to enhance and improve the contrast. In the feature extraction, the Histogram of Oriented Gradients has been applied to extract the image's feature. Finally, in classification two different machine learning techniques (Support Vector Machine and Logistic Regression) have been used to classify COVID-19 and normal patients. The result analysis shows that the SVM achieved the highest accuracy of 96% and provide a better result than logistic regression (92% accuracy).
Naimullah, Syed Irfan Ullah, Arbab Wajid Ullah, Abdus Salam, Muhammad Imad, and Farhat Ullah
Springer International Publishing
Conventional network architectures are not appropriate for the needs of current businesses, carriers, and end-users. Consequently, new evolving network architecture, Software Defined Network can be used to solve a problem as it is more adaptive, dynamic, manageable, and programmable. The SDN architecture control and data planes are isolated, network information and the state are legalized, and the essential network infrastructure is excluded from the applications. However, a network may achieve a point where the computer or network resources restrict the data flow that is controlled according to the bandwidth. In this paper, the custom network topology is created. To observe the performance of Dijkstra's shortest route algorithm in the SDN open-source controllers: RYU and POX to find out the shortest route between source and the destination nodes. The controller's performance is calculated based on the quality-of-service measures, containing throughput and packet delivery ratio in custom network topology under the various workload. The Mininet emulator tool is used to construct the custom network topology. The results are obtained by simulating different QoS requirements in customized topology in Mininet.
Muhammad Abul Hassan, Syed Irfan Ullah, Abdus Salam, Arbab Wajid Ullah, Muhammad Imad, and Farhat Ullah
Institute of Advanced Engineering and Science
<span>Flying Ad-hoc networks are emergent area in Ad-hoc networks evolved from MANETs and VANETs. Small unmanned aerial vehicles (UAVs) are used in FANETs applications and these small UAVs have limited resources while efficiently utilization of these resources is most critical task in real time monitoring of FANETs application. Network consumes its resources in path selection process and data routing from source to destination. Selecting of efficient routing protocol to utilize all available resources played vital role in extending network life time. In this article fisheye state routing (FSR) protocol is implemented in FANET and compare networks performance in term of channel utilization, link utilization vs throughput and packet delivery ratio (PDR) with distance sequence distance vector (DSDV), optimized link state routing (OLSR), adhoc on demand distance vector (AODV), dynamic source routing (DSR) and temperary ordered routing protocol (TORA). Experimental analysis slows that FSR is good in term of PDR (16438 packets delivered), channel utilization (89%) and link vs throughput from the rest of routing protocols after addressing of these problems UAVs resources are efficiently utilized (energy).</span>
Faiza, Syed Irfan ullah, Abdus Salam, Farhat Ullah, Muhammad Imad, and Muhammad Abul Hassan
IEEE
In recent years, the rate of skin diseases is increasing worldwide, skin cancer is defined as the rapid growth of skin cells due to DNA damage which cannot be repaired. It can be harmful and can lead to death if not diagnosed at early stages. The rapid growth of technology, makes it possible to detect different skin diseases at early stages. The impact of rapid technological change on sustainable development in the areas of image processing and machine learning gives an ability to detect early, which increases the probability of survival in the cancer patients. This research primarily focuses on segmentation and classification of the skin lesions from the MRI scan images. Segmentation is carried out in three stages which are pre-processing, segmentation and postprocessing. In the second section, classification is performed using feature extractor and different classifiers. The features are firstly extracted using color, shape, texture component of the skin lesion, and then concatenate the result of all feature distractors. Finally, assign the result of feature distractor to a different classifier to compare most and good classifiers used in this article. The result analysis of the proposed system shows the Naive Bayes, Logistic Regression, and Support Vector Machine classifier provide a better accuracy up to 92%,92%,89.6% respectively among other classifiers.