Improved DF-GAN with Wasserstein Gradient Penalty for Text-to-Image Synthesis Aws Al Tamimi, Ban N. Dhannoon 3rd International Conference on Business Analytics for Technology and Security Icbats 2025, 2025 A novel approach for text-to-image synthesis employs Deep Fusion Generative Adversarial Networks (DFGAN) integrated with the Wasserstein Gradient Penalty (WGAN-GP) to promote training stability and enhance image quality. The model enforces 1-Lipschitz continuity in the discriminator, ensuring smooth gradient flow and mitigating issues such as mode collapse. Evaluations on the Caltech UCSD Birds 200-2011 (CUB) dataset, using metrics like the Inception Score (IS) and Fréchet Inception Distance (FID), demonstrate its capability to generate high-resolution, semantically accurate images that correspond well to textual descriptions. Furthermore, an exploration of various training configurations offers valuable insights for future research in text-to-image synthesis.
Efficient Skin Lesion Segmentation Using U-Net and Mobilenet with Attention Mechanisms Dhamyaa Abd alkareem Farman, Ban N. Dhannoon 3rd International Conference on Business Analytics for Technology and Security Icbats 2025, 2025 Image segmentation is a fundamental and serious medical subject and a necessary step in computer-aided diagnosis systems (CAD). U-Net is the dominant image segmentation design owing to its adaptability, simplified integrated design, and efficacy across all medical imaging modalities. Voluminous modifications of U-Net have been suggested to address the complications made via medical tasks; skin cancer represents a major global health issue, and early precise analysis is essential for enhancing patient outcomes. Skin lesion segmentation can be challenging due to the difference in lesion sizes, brightness fluctuations, texture discrepancies, positional variations, colour differences, and extraneous objects, for instance, air bubbles, hair, or ruler outlines. In this work, we present a deep learning structure based on the U-Net design using a MobileNet model as an encoder to deal with the semantic segmentation of skin images effectively. This approach integrates the pre-trained MobileNet model and attention modules with the UNet, which has a wellorganized architecture. A comprehensive experimentation of our suggested architecture uses Quantitative metrics to evaluate segmentation results, including the Dice coefficient (DS) and intersection over union(IoU). This indicates that the proposed method can discriminate meaningful regions in skin images. The proposed model was assessed using four freely accessible databases: ISIC 2016, ISIC 2017, ISIC 2018, and PH2. The International Skin Imaging Collaboration announced the first three, while The Dermatology Department contributed the fourth dataset at the Pedro Hispano Hospital in Matosinhos, Portugal. The suggested model surpassed U-Net and several contemporary deep-learning segmentation networks for skin lesion segmentation, yielding performance comparable to other networks.
Detailed Cloud Linear Regression Services in Cloud Computing Environment Omer K. Jasim Mohammad, Mohammed E. Seno, Ban N. Dhannoon Informatica Slovenia, 2024 This paper presents a novel cloud-based machine learning framework centered around a linear regression method known as Cloud Linear Regression (CLR). CLR combines elements of cloud technology and machine learning principles. Furthermore, it explores the connection between cloud task scheduling, distribution, and machine learning methodologies, showcasing how linear regression techniques play a pivotal role in enhancing the cloud environment. CLR demonstrated it is effectiveness in dealing with expansive environments that have big data by exhibiting high thorough mining for the best resource predictive accuracy and response times, it has been applied to three scenarios for the best CPU accuracy utilization of the prediction which was (45 %), (53.44 %), and (59.81%) respectively. Moreover, CLR offers an efficient remedy for managing resources, including task scheduling, provisioning, allocation, and ensuring availability. CLR obtained the highest performance of (40%) with multitasking resources, (72%) with Memory utilization, (90% with logical Disk utilization, and (30 %) with Bandwidth utilization.
Word embedding for detecting cyberbullying based on recurrent neural networks Noor Haydar Shaker, Ban N. Dhannoon Iaes International Journal of Artificial Intelligence, 2024 <span lang="EN-US">The phenomenon of cyberbullying has spread and has become one of the biggest problems facing users of social media sites and generated significant adverse effects on society and the victim in particular. Finding appropriate solutions to detect and reduce cyberbullying has become necessary to mitigate its negative impacts on society and the victim. Twitter comments on two datasets are used to detect cyberbullying, the first dataset was the Arabic cyberbullying dataset, and the second was the English cyberbullying dataset. Three different pre-trained global vectors (GloVe) corpora with different dimensions were used on the original and preprocessed datasets to represent the words. Recurrent neural networks (RNN), long short-term memory (LSTM), Bidirectional LSTM (BiLSTM), gated recurrent unit (GRU), and Bidirectional GRU (BiGRU) classifiers utilized, evaluated and compared. The GRU outperform other classifiers on both datasets; its accuracy on the Arabic cyberbullying dataset using the Arabic GloVe corpus of dimension equal to 256D is 87.83%, while the accuracy on the English datasets using 100 D pre-trained GloVe corpus is 93.38%.</span>
Modified Bag of Visual Words Model for Image Classification Computer Science Department, College of Science, Al-Nahrain University, Baghdad, Iraq, Zainab N. Sultani, Ban N. Dhannoon, Computer Science Department, College of Science, Al-Nahrain University, Baghdad, Iraq Al Nahrain Journal of Science, 2021
Color model based convolutional neural network for image spam classification Department of Computer Science, College of Science, Al-Nahrain University, Baghdad, Iraq, Ahmad Mahdi Salih, Ban Nadeem Dhannoon, Department of Computer Science, College of Science, Al-Nahrain University, Baghdad, Iraq Al Nahrain Journal of Science, 2020
A feature selection approach using binary Firefly Algorithm for network intrusion detection system Arpn Journal of Engineering and Applied Sciences, 2018
An indirect MSB data hiding technique Life Science Journal, 2013
