Computer Engineering Commons^™

Long-Tail Object Detection Method In Transportation Scene Images Based On Improved Rt-Detr, Mingjun Deng, Peihua Nie, Junhua Guo, Ningbo Gao

Journal of System Simulation

Abstract: Detecting rare objects in transportation environments is crucial for the safety and reliability of autonomous driving systems. To address the challenges of limited detection capability on rare objects that appear only a few times during training, we propose an improved long-tailed object detection method for traffic scene images named TF-DETR (tail free DETR) based on RT-DETR model. First, the lightweight CSP-EDA (CSP-efficient depthwise aggregation) was constructed to reconstruct the backbone network. Its core unit, the efficient depthwise aggregation, integrated the pooled-transpose attention and convolutional gated linear unit to enhance feature representations for tail-class samples. Second, the spatially-adaptive dynamic cognitive …

Hybrid Attention For Low-Overlap Point Cloud Registration, Huaijun Wang, Zonghua Yu, Chenxin Xu, Junhuai Li, Chenzhi Li, Junjie Yin

Journal of System Simulation

Abstract: To address the significant registration errors caused by large initial pose deviations in low overlap ratio point cloud, this paper presents a two-stage optimization method that integrates selfattention and cross-attention mechanisms. Orientation feature representation is enhanced using equivariant self-attention, semantic alignment is achieved through invariant cross-attention, and rotational equivariance is modeled with an octahedral group for coarse registration. Various attention variants are employed to aggregate features along the anchor dimension, and a local-to-global module is utilized for fine registration. On the 3DLoMatch dataset, the proposed method achieves a registration recall of 77.0% and an inlier ratio of 44.3%. Even …

Dynamic Focal Loss Adjustment For Railway Defect Detection, Mehmet Koç, Ridvan Özdemi̇r, Ömer Gerek

Turkish Journal of Electrical Engineering and Computer Sciences

Railway infrastructure is critical to the safe and efficient operation of transportation systems, and the early detection of defects is essential for preventing catastrophic failures. Automated defect detection methods are therefore crucial for maintaining continuous safety while reducing maintenance costs. Although Focal Loss is widely used in object detection under class-imbalanced conditions, its fixed α parameter may limit its effectiveness in detecting rare defects. In this study, we propose an adaptive α-tuned Focal Loss approach that dynamically adjusts class weights based on average precision (AP) values. By iteratively optimizing α without relying on gradient-based optimization, the proposed method improves the …

Semantic-Guided Diffusion Autoencoder For Medical Image Synthesis And Representation Learning, Neda Yousefi, Abbas Koochari, Azadeh Zamanifar

Makara Journal of Technology

The development of robust machine learning models for medical diagnostics is constrained by the scarcity of large, annotated imaging datasets, a challenge rooted in high acquisition costs and stringent privacy regulations. To address this, generative approaches for data synthesis have emerged as an essential tool, and this study explores the use of diffusion models to generate medical images. This study introduces a novel Diffusion Autoencoder Model (DAM) integrated with a semantic encoder framework to generate interpretable and decodable image representations. Our DAM learns disentangled latent features via its encoder, which then conditions a diffusion probabilistic model acting as the decoder, …

Deep Multimodal Sensor Fusion For Autonomous Driving: A Real-World Deployment On Little Blue, Raphael T. Pamie-George

Electrical & Computer Engineering Theses & Dissertations

Accurate perception and control systems are essential for autonomous vehicles to safely interpret and navigate complex environments. Individual sensing modalities such as cameras, LiDAR, and inertial measurement units (IMUs) provide valuable environmental and vehicle state information; however, each sensor exhibits inherent limitations when used independently. Camera performance can degrade under varying lighting conditions, LiDAR measurements may be sparse or affected by environmental noise, and IMU measurements are susceptible to drift over time. Little Blue, a research autonomous vehicle platform from ODU (Old Dominion University), previously relied primarily on camera-based data for neural network training, limiting the system’s ability to exploit …

Eyolov8-Msaff: An Enhanced Object Detection Algorithm With Multi-Scale Attention And Feature Fusion For Small And Dense Object Detection In Uav Applications, Mohammed Hasan Mutar, Morteza Valizadeh, Alaa Hussein Abdulaal, Mehdi Chehel Amirani

Iraqi Journal for Computer Science and Mathematics

The proliferation of unmanned aerial vehicles (UAVs) has necessitated the development of sophisticated object detection algorithms capable of handling the unique challenges posed by aerial imagery. Traditional detection methods often struggle with small object sizes, dense distributions, and complex backgrounds characteristic of UAV-captured scenes. This research presents EYOLOv8-MSAFF (Enhanced YOLOv8 with Multi-Scale Attention and Feature Fusion), a novel deep learning architecture specifically engineered for superior performance in UAV-based object detection tasks. The proposed methodology integrates four innovative components: a Hybrid Spatial-Channel Attention Mechanism (HSCAM) that processes attention information in parallel rather than sequentially, an Adaptive Multi-Scale Feature Fusion Module (AMSFFM) …

A Novel Hierarchical Multi-Scale Attention Mechanism For Enhanced Yolov8 Performance In Unmanned Aerial Vehicle-Based Small Object Detection, Mohammed Hasan Mutar, Morteza Valizadeh, Alaa Hussein Abdulaal, Mehdi Chehel Amirani

Iraqi Journal for Computer Science and Mathematics

Object detection in Unmanned Aerial Vehicle (UAV) images presents significant challenges due to the prevalence of small and densely packed objects, as well as variations in scale, orientation, and lighting conditions. This paper introduces a novel object detection algorithm, Hierarchical Multi-Scale Attention YOLO (HMSA-YOLO), which is an improved version of YOLOv8 designed to address these challenges. The proposed method incorporates a novel Hierarchical Multi-Scale Attention (HMSA) module, a Bidirectional Feature Pyramid Network (BiFPN) for enhanced feature fusion, a modified loss function, and an adaptive anchor optimization technique. The HMSA module effectively captures both channel and spatial dependencies at multiple scales, …

Scalefusion: Hierarchical Feature Aggregation For Unified Multiscale Object Detection, Ayşenur Yaylaci, Berru Kaya, Mehmet Kiliçarslan

Turkish Journal of Electrical Engineering and Computer Sciences

Detecting objects across a wide range of scales, particularly small ones, remains a significant challenge in computer vision. Existing methods often improve small object detection at the cost of performance on larger objects or introduce significant computational overhead through external techniques like image slicing. This paper introduces ScaleFusion, a novel, unified, end-to-end object detection architecture designed to provide robust performance across all scales within a single model. The core of our approach is a hierarchical feature aggregation strategy structured like a tree. ScaleFusion processes an image by running a shared backbone network only on fine-grained patches at the lowest level …

Video Compression Optimization Techniques Using Artificial Intelligence: Review/Review Article, Amal Abbas Kadhim, Wedad Abdul Khuder Naser, Nada Abdulkareem Hameed

Al-Esraa University College Journal for Engineering Sciences

Compressing video is an important and essential function in today's multimedia systems as it enables the efficient storage and transmission of large volumes of video data. The proliferation of high-resolution videos that are being used in various application areas such as video streaming, video conferencing, surveillance, and autonomous systems has caused the strong demand for more efficient compression algorithms. This paper presents an in-depth review of video compression techniques with particular focus on AI methods. It also discusses traditional video coding standards including H. 264/AVC, H. 265/HEVC, and AV1, including motion estimation, transform coding quantization entropy coding, and rate-distortion optimization. …

Deep Learning Approaches For Ctenophore Identification And Tracking, Anagha Bharadwaj

Theses

Ctenophores are translucent marine organisms with nearly invisible tentacles and pose significant challenges due to their transparent morphology and ambiguous structural features. This research addresses the classification and tracking of these organisms and evaluates the performance of current computer vision models under sparse-data environments.

A dataset from the NJIT Life History Lab consisting of microscopic laboratory videos and photographs of different growth stages is used to train and assess a number of convolutional neural network designs, including VGG16, ResNet, BioCLIP2, YOLO, and DeepLabCut. Additionally, a web-based interface is developed to evaluate expert-labeled ground truth with the model's performance.

The findings …

A Descriptive Analysis Of Plant Leaf Disease Detection Using Machine Learning And Deep Learning Models: A Systematic Review, Arzoo Chamoli, Anuj Kumar

Turkish Journal of Electrical Engineering and Computer Sciences

Plant leaf disease detection (PLDD) is a growing active research area with burgeoning practical applications across various sectors such as agricultural monitoring, food security, and environmental conservation. Accurate segmentation and classification of plant leaf diseases remains a key challenge in the field of plant leaf disease prediction. The challenge demands automated methods for the plant disease identification because it needs to develop better crop management systems, which will boost agricultural production. In this article, we provide a systematic review of various machine learning (ML) and deep learning (DL) methods extensively used for PLDD. The review strategy follows a formal protocol, …

Multimodal Aerial Image-Based Ground Object Detection And Classification Using Yolo, Erik Nielsen

All Theses

Object detection in unmanned aerial vehicles (UAVs) present a unique challenge due to small object sizes, varying viewpoints, and changing environmental conditions. These challenges are exacerbated when operating during daytime and nighttime scenarios where illumination differences can heavily impact detection performance. This work is motivated by military object detection applications where the ability to reliably identify small objects such as landmines or unexploded ordnance from aerial imagery presents a critical safety need and a significant technical challenge. In this paper, we investigate object detection using both visible (RGB) and infrared (IR) imagery to improve robustness and reliability across diverse operating …

Enhancing Lane Detection In Autonomous Vehicles Using Data Augmentation For Adverse Environmental Conditions, Rutvikkumar Dave

Electronic Theses and Dissertations 2020 - Present

To make sure that self-driving and connected automobile technologies are safe and work well, it’s really important that they can correctly identify lanes. But lane detection Algorithms typically have a hard time working well when the weather is bad, such when it rains, fogs, or goes too fast. The circumstances cause visual distortions that make existing computer vision systems less reliable, which makes it harder requires autonomous navigation systems to work well. This paper introduces a comprehensive lane detection system that integrates synthetic Weather-informed data augmentation combined with a Weather-aware Temporal Lane Detection Network (WTLDNet) to make it easier for …

Modelling Method Of Unmanned Vehicle Dynamics Based On Neural Network, Jun Wang, Min Liu, Xiaochuan Zhang, Yishan Ding, Juhui Feng, Ye Zhuang

Journal of System Simulation

Abstract: To address the challenges of high data acquisition costs of test data on dynamic characteristics between tires and soft terrain and low speed of numerical calculation for unmanned vehicles in complex terrestrial environments, a modeling method of unmanned vehicle dynamics based on a neural network was proposed. Tire-terrain contact dynamics models were built by using discrete element method (DEM) simulations for tire-terrain contact and experimental data, thereby creating a dataset of tire contact forces for various tire materials in terrestrial environments. The neural network was applied to regressively learn the dataset, and a nonlinear neural network tire model was …

Large-Scale File Fragment Classification Via Multi-View Learning, Samuel Hildebrand

LSU Master's Theses

File reassembly is one of the most fundamental tasks in digital forensics, enabling recovery of data from potentially damaged storage media even when file system metadata is unavailable. This thesis reviews more than two decades of work in the realm of file carving, with a particular focus on fragmented file carving, which remains a focus of research, and file fragment classification, a principal component of fragmented file carving. This thesis serves a literature review of both file carving and fragmented file carving, surveys the massive amounts of data needed for the task of fragment classification and the datasets that serve …

Simulation And Multi-Perspective Recognition Algorithm For Typical Trajectory Shapes, Xuejian Feng, Han Ding, Yiqi Tong, Chaoying Huo, Yanjin Zhang

Journal of System Simulation

Abstract: Current trajectory simulation methods inadequately address geometric shape features and kinematic properties of the target trajectory. To bridge this gap, a target trajectory shape simulation algorithm based on kinematic laws was proposed. The polar coordinate equations and curvature equations of multiple trajectories were integrated. The aircraft state parameters were solved by combining kinematic equations. Angular Gaussian noise was introduced to enhance trajectory diversity and authenticity. Additionally, a multi-perspective trajectory shape recognition algorithm was designed, which could effectively integrate image and sequential multi-modal features by adopting a multilayer perceptron, enabling precise trajectory shape recognition. Experimental results demonstrate that the proposed …

Research On Visual Place Recognition Algorithms For Complex Urban Environments, Peijin Liu, Minxin Zhang, Lin He, Yige Sun, Tingqi Su

Journal of System Simulation

Abstract: Dynamic factors such as traffic flow and crowd density in complex urban environments reduce the accuracy of visual place recognition (VPR) algorithms. To solve these problems, a semantic-guided visual place recognition (SG-VPR) algorithm was proposed. A semantic-guided feature suppression module was designed. A semantic-guided module and feature suppression layer were constructed to reduce the dynamic object interference and more accurately extract the key static features. An adaptive triplet margin loss function (ATML) was proposed by improving the traditional triplet margin loss. The margins were adaptively adjusted according to the sample distribution, solving the problem of suboptimal solution convergence …

Spatio-Temporal Swin Transformer-Based Flow-Solid Coupling Interaction Sequence Image Prediction Network, Changjun Zou, Zhiyu Ge, Chenxi Zhong

Journal of System Simulation

Abstract: To address limitations in modeling long-term dependencies and multi-scale features in fluidstructure interaction scenarios, a spatiotemporal deep learning model (SwinLSTM) integrating ConvLSTM and Swin Transformer is proposed. The model employs a gated spatiotemporal attention mechanism that dynamically embeds Swin Transformer's window-based multi-head self-attention into ConvLSTM's output gate, enabling adaptive temporal-spatial feature coupling, and designs a multi-level ConvLSTM framework to hierarchically capture complex spatiotemporal correlations. Experiments on a self-built fluid-interaction dataset show that our method achieves the highest PSNR and leading SSIM scores, with superior performance in preserving vortex details and boundary consistency. This work provides an efficient solution …

Ai For Life Sciences: From Geometric Protein Modeling To Multimodal Drug Design, Feng Jiang

Computer Science and Engineering Dissertations

Predicting biomolecular interactions, from immune recognition to drug–target binding, is a central problem in the life sciences and computational drug discovery. Deep learning has advanced this area, yet three challenges persist: the topology of large, highly imbalanced interaction networks; structural noise in computationally predicted protein models; and the integration of multimodal information such as functional text and taxonomic annotations. This dissertation develops a coherent set of models spanning immune complex prediction and small-molecule drug design: graph learning that addresses network topology and severe class imbalance; a noise-tolerant method that fuses predicted structures with evolutionary sequence features; and multimodal representation learning …

Deep Learning Frameworks For Functional Genomics: Lncrna–Mirna Interactions, Dose-Specific Graph Modeling, And Complete Graph-Based Lncrna Functional Annotation Using Cross-Layer Attention, Sachit Satyal

Computer Science and Engineering Dissertations

Functional genomics seeks to understand how genes and regulatory elements orchestrate biological processes and adapt under diverse conditions. Two critical challenges in this field are deciphering the regulatory roles of long noncoding RNAs (lncRNAs) and modeling cellular transcriptional dynamics under therapeutic perturbations. While lncRNAs are increasingly recognized as key regulators in health and disease, their functions remain incompletely characterized. Concurrently, advances in single-cell transcriptomics reveal how gene and pathway activities rewire across drug dosage gradients, highlighting the need for computational frameworks that are predictive, interpretable, and capable of integrating molecular interactions with dynamic cellular states. This dissertation develops a unified …

Applying Machine Learning Techniques For Early Detection Of Cyber Attacks On Iot Devices, Noor Adnan Allamy

Al-Esraa University College Journal for Engineering Sciences

This research designs, implements, and evaluates a machine learning-based framework for the early detection of cyber attacks targeting Internet of Things (IoT) devices, with a specific focus on the context and challenges present in Iraq. The study conducts a comparative analysis of three supervised learning algorithms—Support Vector Machine (SVM), Random Forest (RF), and Deep Neural Networks (DNN)—using a combination of benchmark datasets (NSL-KDD, CIC-IDS-2017, Bot-IoT) and a synthesized dataset adapted to simulate the Iraqi threat landscape. Key performance metrics, including accuracy, precision, recall, and F1-score, were used for evaluation. The proposed Random Forest model demonstrated superior performance, achieving an accuracy …

Enhanced Generative Convolutional Networks: A Hybrid Algorithm For Refinement Video Classification, Dalal Thair Mahjoub, Hala Bahjat Abdulwahab, Kesra Nermend

Journal of Soft Computing and Computer Applications

Video classification is a vital area of research due to the growing volume of video content in various applications. Accurate category across various resolutions poses challenges, which include adapting to scaling, resizing, and compression. Therefore, this paper introduces an innovative Generative Convolutional Network (GCN) set of rules tailored for multi-resolution video classes. The proposed GCN model utilizes Convolutional Neural Networks (CNNs) combined with generative modeling to enhance the extraction of functions across varying video resolutions, which is crucial for maintaining class robustness in the face of common video adjustments, such as scaling, resizing, and compression. In contrast, traditional fashions frequently …

Review Of Video Steganography By Using Deep Learning Methods: Datasets, Techniques, And Evaluations, Noor Fahem Sahib, Soukaena Hassan Hashem, Ekhlas Falih Naser

Journal of Soft Computing and Computer Applications

The growing prevalence of cyber threats, including fraud and attacks, has intensified the demand for secure methods of safeguarding confidential information exchanged between users. As telecommunications increasingly rely on multimedia data, video steganography has become a prominent technique to address these concerns. By embedding sensitive data within video files, this approach enhances protection against unauthorized access and common internet-based attacks, offering a robust layer of security in an era of escalating digital risks. With the introduction of Deep Learning (DL) steganography methods recently, video steganography can be defined as a rapidly developing subject within information security. This study provides a …

Research On Infrared And Visible Light Fusion Method Based On Resnet-50 And Laplacian Filtering, Xiao Wang, Xiangyang Li, Feng Liang, Zhili Zhang

Journal of System Simulation

Abstract: In order to solve the problem that existing infrared and visible light image fusion techniques often suffer from artifacts caused by insufficient contrast, spectral distortion, and high computational complexity, a fusion framework based on ResNet-50 and Laplacian filtering was proposed. ResNet-50 was used to extract shallow and deep features, followed by multi-scale feature fusion. Laplacian filtering was applied to optimize feature information, and an automatic discriminator was introduced to further improve the fusion effect. Simulation results show that, compared with comparison algorithms, the proposed method achieves an average increase of 2.71% and 2.16% in information entropy, 5.98% and …

Green Ai-Enhanced Deep Learning Model For Breast Cancer Detection And Classification In Mammography Images: Bc-Net-512, Nesma Abd El-Mawla, Mohamed A. Berbar, Nawal A. El-Fishawy, Mohamed A. El-Rashidy

Mansoura Engineering Journal

This study champions a sustainable approach for developing a Deep Learning (DL) model for medical image analysis, specifically focusing on breast cancer (BC) detection in mammograms. By prioritizing low-computing algorithms to achieve high diagnostic accuracy while minimizing the model's environmental footprint, that aligns with the principles of Green AI. In this paper, an innovative architecture called BC-Net-512 was constructed for the classification of BC mammography. It is composed of lightweight Convolutional Neural Network (CNN) blocks for texture, density, and structure feature extraction and detection, a thin, fully connected layer for learning complex patterns and correlations in the extracted features, and …

A Comprehensive Review Of Dental Diseases Detection And Classification Based On Artificial Intelligence Techniques, Nermeen N. Noaman, Yasmin M. Alsakar, Naira E. Elazab, Waleed M. Mohamed, Mohamed E. Ezzat, Mohammed M. Elmogy

Mansoura Engineering Journal

In dentistry, many diseases, such as gum, cavities, and oral cancer, affect people of all ages. Early treatment and diagnosis are crucial for minimizing dental diseases' effect on overall health and saving money in the long run. Traditional dental diagnosis methods, such as manual probing and visual inspection, are time-consuming and can be subject to human errors. Hence, a computer-aided diagnosis system based on computer vision and artificial intelligence (AI) techniques is needed. The considerable progress in computer vision and AI techniques offers many possibilities in dental diagnosis based on dental X-ray imaging modalities. Dental X-rays are used to diagnose …

Improving Road Safety Through Multimodal Deep Learning For Driver Drowsiness Detection, Hadel A. Hussain, Mohammed A. Subhi, Ahmed S. Al Tmeme, Ahmed D. Radhi, Marwan Ali Albahar

Iraqi Journal for Computer Science and Mathematics

One of the most common causes of road accidents globally is driver drowsiness and it needs solutions that are reliable and can be applicable in numerous real-life situations. We present this paper with the aim of developing a deep-learning system that is capable of reliably detecting drowsiness in diverse and varied conditions across different drivers, environments, and sensor types. Our system is known as Multimodal Attention Network (MMAN), which combines information of eye and head movement, heart-rate and breathing pattern, and vehicle-dynamics signal. MMAN has a gradient-reversal layer that enables the layer to be domain-adaptive such that it does not …

A Feature-Free Deep Learning Approach For Midair Hand Gesture Recognition From Surface Electromyogram (Semg) Data, Yasir Altaf, Abdul Wahid, Mudasir Manzoor Kirmani

Turkish Journal of Electrical Engineering and Computer Sciences

Midair hand gesture recognition plays a crucial role in applications such as sign language recognition and human-computer interaction, particularly for supporting individuals with partial or complete hearing loss. However, recognizing gestures in midair remains challenging due to the rapid and complex nature of hand movements. To address this, noninvasive techniques like surface electromyography (sEMG)—which captures muscle activity through sensors placed on the skin—have gained attention. sEMG provides rich time-series data that reflect both spatial and temporal muscle dynamics. In this study, we propose a deep learning architecture that combines convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to classify …

Seeing What Matters: Safety-Critical Semantic Segmentation Via Transfer Learning On Construction Sites, Obiora J. Odugu

LSU Master's Theses

Autonomous robots are increasingly deployed on construction sites for tasks such as progress monitoring, inspection, and safety assessment. For these robots to operate effectively, they must perceive and interpret complex, dynamic environments populated by workers, machinery, and unstructured terrain. Achieving reliable perception depends on high performing semantic segmentation models trained on large volumes of annotated data—an expensive and logistically challenging requirement in construction due to privacy restrictions, variable site access, and slow digitalization. This research addresses the challenge of limited labeled data by investigating transfer learning as a label-efficient approach for construction-site segmentation. Specifically, it explores whether road construction imagery—abundant …

Soft Sensor Modeling Based On Improved Transformer In Dual-Stream Framework, Hao Gu, Jiayu Wang, Weili Xiong

Journal of System Simulation

Abstract: Industrial process information is highly nonlinear and dynamic, with long-term dependencies between data, making it difficult to adequately extract time-series features. To address this issue, an improved Transformer-based soft sensor model in a dual-stream framework was proposed. The data were segmented and expanded. The features were extracted in parallel using a dual-stream structure combining a convolutional neural network with a self-attention mechanism and the improved Transformer model. The dual-stream features were fused for soft sensor regression. Residual connections were further introduced to accelerate the convergence speed of the model, and an orthogonal random features-based improved multi-head attention mechanism was …