
Academic Excellence • Open Access • Global Reach
UU Academic Repository
Discover, preserve, and share the scholarly achievements of Uttara University’s faculty, researchers, and students — accessible to the world.
Preserved Research
Global Open Access
Permanent Archiving
Recent Submissions
Smart Vehicle Monitoring System (SVMS) for Road Safety: A Prototype for Monitoring Drowsiness, Alcohol, and Overload
(2025 IEEE International Conference on Quantum Photonics, Artificial Intelligence, and Networking, QPAIN 2025 Conference Paper, 2025-09-22) Al Mamun Mizan; A. Z. M. Tahmidul Kabir; Nadim Zinnurayen; Tawsif Abrar; Akib Jawad Ta-sin; Mahfuzar
The principal intent behind the titled project is
to make a smart vehicle management system that will prompt
us to reduce traffic accidents. There are three major elements
in this prototype of the smart vehicle management system.
Firstly, there is a drowsiness detector, which will identify the
drowsiness of the driver throughout driving time. Secondly, an
alcohol detector will trace if there is any presence of alcohol on
the body of that particular driver. Lastly, there is an overload
detector, which will show if the vehicle is overloaded, or not.
Low-Cost Integrated Airbag Safety System in IoT-Based Smart Wearables
(2025 IEEE International Conference on Quantum Photonics, Artificial Intelligence, and Networking, QPAIN 2025 Conference Paper, 2025-08-02) Arnab Sarke; Md. Mahbub-E-Elahi Prodhan; Samiu Javed Islam; Sakib Bin Mizan; Risul-Ur-Rashid; Shamim Forhad; Tarifuzzaman Riyad; Md. Maidul Islam Shehab; Mohammad Alif Arman
Drowning is one of the major causes of accidental
death around the world, especially where there is no access to a
rapid rescue system. The conventional life jackets provide
buoyancy but are insufficient in critical situations when the
victim is unconscious or is not able to call for help. To overcome
this shortcoming, this study presents an independent wearable
safety device that can sense underwater distress and activate
immediate rescue services. The system is set to automatically
react to real-time physiological and environmental conditions
without any user intervention for personal protection and child
safety. It combines a live server-based alerting system and dualfunction design, unlike the current solutions, but cost effective
for widespread availability. Currently available technologies
tend to be limited by expense, single use design, or absence of
communication capability, and thus are not usable to extensive
application. This study presents an innovation that fuses smart
sensing, real-time notifications, and low cost which sets it apart
from available commercial systems. In comparison with the
similar easily accessible alternatives, it estimates to be around
47% cheaper without compromising performance. It has been
field tested and assessed under a wide range of weights to ensure
efficiency. The end product is a working, scalable, and costeffective solution that can easily lower drowning rates, especially
in the aquatic environment with limited resources.
Heart Attack Prediction: A Comparative Analysis of Supervised Machine Learning Algorithms for Early Detection and Risk Stratification
(2025 IEEE International Conference on Quantum Photonics, Artificial Intelligence, and Networking, QPAIN 2025 Conference Pape, 2025-08-02) Shamim Forhad; Tanha Mim; Zarin Mosarat; Md. Siam Taluckder; Abdullah Al Masum; Md Tawfiqul Islam; Md. Riazat Kabir Shuvo
—Heart attacks are a prominent source of
morbidity and mortality globally, demanding the development
of precise and efficient predictive models for early identification
and risk stratification. This paper gives a complete review of
supervised machine learning algorithms to measure their
efficacy in predicting MI risk, allowing for timely medical
intervention. A wide range of clinical data were used to train
and verify six classification models. Based on accuracy,
precision, recall, and F1-score, LR had the best prediction
performance with an accuracy of 85.71%, followed by KNN
(84.62%), SVM (80.22%), RF (79.12%), XGBoost (79.12%), and
DT (71.43%). The performance of LR demonstrates its practical
applicability and interpretability for clinical use, making it a
viable candidate for clinical deployment. These findings
emphasize the potential of machine learning-driven technologies
in improving MI prediction, therefore leading to increased early
diagnosis and individualized patient treatment. The study
further emphasizes the need for integrating transparent and
interpretable ML models into healthcare systems to facilitate
informed clinical decision-making and optimize patient
outcomes.
A Framework for Accurate and Efficient Image Captioning by Fusing Fine-Tuned YOLOv8 and LLMs
(2025 IEEE International Conference on Quantum Photonics, Artificial Intelligence, and Networking, QPAIN 2025 Conference Paper, 2024-02-05) Dabbrata Das; Md Rishadul Bayesh; Mohammad Amanul Islam; Md. Asifur Rahman
Automatic image captioning is essential for generating natural language descriptions by extracting meaningful features and understanding contextual relationships
in images. While traditional methods like CNN-RNN models struggle with computational complexity and spatial
awareness, multimodal Large Language Models (LLMs)
offer an alternative but often lack object precision and are
computationally expensive. In this work, we propose a
novel image captioning framework that combines a finetuned YOLOv8 model with an LLM for efficient and accurate caption generation. YOLOv8 detects objects, extracts their names, confidence scores, and bounding box
coordinates, which are filtered based on confidence levels
above 0.5 before being passed to the LLM. This integration results in richer, more contextually accurate captions
with lower inference time compared to existing methods.
We evaluate our approach against multimodal LLMs and
CNN-RNN models, demonstrating that it significantly improves efficiency while maintaining high caption quality.
This method provides a promising solution for real-time
applications, offering faster and more reliable image captioning for systems such as autonomous technologies and
content generation.
Investigation of thermo-physiological comfort characteristics of knitted fabrics made from regenerated cellulose and nylon blended mélange yarn
(Research Journal of Textile and Apparel, 2024-08-08) M. Fazley Elahi; Md Anwar Hossain; Md. Atiqur Rahman; M. Rayhan Siddique; Mohammed Rafin Bhuiyan; Habibur Rahman; Md. Reajul Islam; Mohammad Rafiqur Rashid
This study investigates the comfort characteristics of mélange fabrics made from Tencel, Modal, Ecovero, and Bamboo
fibers, each blended with 20% virgin nylon. The comfort properties of regenerated cellulosic fiber-blended mélange
fabrics were evaluated by analyzing various parameters, including moisture management, breathability, hand feel, and
thermal properties. Dimensional stability, spirality, pilling, and bursting strength were also examined to assess the
physical performance of the fabrics. The experimental results revealed that Tencel-blended mélange fabric
demonstrated superior comfort properties, offering better moisture management, breathability, hand feel, and thermal
properties. Additionally, Tencel-blended mélange fabric showed higher bursting strength and optimal levels of
dimensional stability and spirality compared to other regenerated cellulosic fiber-blended mélange fabrics. These
findings underscore the suitability of Tencel-blended mélange fabrics for applications requiring excellent breathability,
moisture management, and durability. The comprehensive analysis of various fabric blends provides valuable insights for
optimizing fabric properties for specific end uses.
