EXTERNAL PROFILES
Assistant Professor
Md. Sabbir Ejaz
Md. Sabbir Ejaz is an Assistant Professor in the Department of Information and Communication Engineering at Noakhali Science and Technology University (NSTU), Bangladesh. His research focuses on machine learning, computer vision, deep learning, and medical image analysis. He actively engages in teaching, research supervision, and student mentoring.
Information and Communication Engineering
BIOGRAPHY
Md. Sabbir Ejaz is an Assistant Professor in the Department of Information and Communication Engineering at Noakhali Science and Technology University (NSTU), Bangladesh. Previously, he served as a Lecturer in the Department of Computer Science and Engineering at Bangladesh Army University of Engineering & Technology (BAUET). He completed both his B.Sc. and M.Sc. in Computer Science and Engineering from Rajshahi University of Engineering and Technology (RUET). His research focuses on machine learning, deep learning, computer vision, medical image analysis, and explainable artificial intelligence, with publications in international journals and conferences. Alongside teaching and research, he actively mentors students, supervises research projects, and contributes to curriculum development, programming clubs, innovation activities, and other student-led extracurricular initiatives.
RESEARCH INTERESTS
2017 - 2020
MSc in Engineering
Computer Science and Engineering
Rajshahi University of Engineering and Technology
2010 - 2016
BSc in Engineering
Computer Science and Engineering
Rajshahi University of Engineering and Technology
Last updated on 2026-07-06 06:57:06
Lecturer
Information and Communication Engineering, Noakhali Science and Technology University
2022-01-24 to
2024-01-23
2024-01-23
Lecturer
Computer Science and Engineering, Bangladesh Army University of Engineering and Technology
2016-08-10 to
2022-01-23
2022-01-23
Assistant Professor
Information and Communication Engineering, Noakhali Science and Technology University
2024-01-24 to
No Training information...
Last Updated: N/A
2023-08-25
Organizing Your Research with ENDNOTE
Description: Enter Training Description
Organization: Bangladesh Students' Union Malaysia
2023-07-15
Project Management
Description: Enter Training Description
Organization: Bangladesh Students' Union Malaysia
2022-08-20
Big Data Analytics
Description: Enter Training Description
Organization: YouthBees
2022-07-17
Improving Writing Skills with Better Impact
Description: Enter Training Description
Organization: Bangladesh Students' Union Malaysia
2022-05-21
Arduino Workshop for Beginners
Description: Enter Training Description
Organization: RoboDoc Limited
2022-03-23
Workshop on Ambient Intelligence: Adaptivity by using artificial intelligence, machine learning and biometrics in worldwide cloud-based environments
Description: Enter Training Description
Organization: IIUC and BUBT
2022-03-16
Python
Description: Enter Training Description
Organization: East West University
2019-07-09
Embedded System Design on Zynq System-on-Chip Architectures
Description: Enter Training Description
Organization: University of Rajshahi
2016-10-09
Method of Instruction
Description: Enter Training Description
Organization: Bangladesh Army University of Engineering and Technology
Last Updated: 2025-10-20 13:24:58
AWARDS AND ACHIEVEMENTS
1
Best Paper Award
Green University
2025
Journal
A Comprehensive Study on the Automatic Identification of Diabetic Retinopathy
Authors: Ejaz, Md and Mahee, Nusrat and Rashid, Md and Faruq, Md
Journal: International Journal of Research and Scientific Innovation
2025
Journal
Identification of Diabetic Retinopathy Using Deep Learning and Ensemble Model Approach
Authors: Nusrat Mahee, Md. Sabbir Ejaz
Journal: Journal of Undergraduate Research International
DOI: 10.64589/juri/214968
2023
Journal
Facial recognition-based entry system for student residence halls: Enhancing security and accessibility
Authors: Ejaz, Md Sabbir and Debnath, Sourav and Hasan, Mohammad Kamrul and Alam, Md Mahbubul
Journal: Asian Journal of Research in Computer Science
2023
Journal
Perceived Factors Analysis for Depression and Suicidal Ideation among Bangladeshi University Students Using Association Algorithm
Authors: Alam, Md. Mahbubul and Ejaz, Md and Hasan, Mohammad Kamrul and Khan, Dr. Md. Ashikur
Journal: Asian Journal of Research in Computer Science
2023
Journal
A Review on Malicious URLs Detection Using Machine Learning Methods
Authors: Tabassum, Tasfia and Alam, Md. Mahbubul and Ejaz, Md and Hasan, Mohammad Kamrul
Journal: Journal of Engineering Research and Reports
2023
Journal
Classifying Bengali Newspaper Headlines with Advanced Deep Learning Models: LSTM, Bi-LSTM, and Bi-GRU Approaches
Authors: Hasan, Mohammad Kamrul and Islam, Shoeb and Ejaz, Md and Alam, Md. Mahbubul and Mahamud, Nahid and Rafin, Tanvir
Journal: Asian Journal of Research in Computer Science
2024
Conference
A Smart Shopping Assistant at Super Shop for Visually Challenged Persons Utilizing Machine Learning Approach
Authors: Saha, Katha and Ejaz, Md Sabbir and Debnath, Sourav
2022
Conference
Early identification of diabetic retinopathy using deep learning model: A survey
Authors: Shrabony, Umma Kulsum and Ejaz, Md Sabbir
2020
Conference
Recognition of Bengali Sign Language using Novel Deep Convolutional Neural Network
Authors: Hossein, Md and Ejaz, Md
2019
Conference
Implementation of principal component analysis on masked and non-masked face recognition
Authors: Ejaz, Md Sabbir and Islam, Md Rabiul and Sifatullah, Md and Sarker, Ananya
2017
Conference
Performance Comparison of Partition Based Clustering Algorithms on Iris Image Preprocessing
Authors: Ejaz, Md Sabbir and Hossain, Md Ali and Matin, Abdul and Ahmed, Md Tanvir
2024-07-01
to
2025-06-30
Research
An Early Automated Screening System for the Diagnosis of Diabetic Retinopathy Using Deep Learning Architectures
Abstract: One of the most prevalent complications of diabetes and a leading cause of preventable blindness in the world is DR. This paper examines the history of the DR detection methods, the transition of traditional image-processing algorithms to the state-of-the-art deep learning and ensemble models. This study employed MESSIDOR dataset; it includes retinal fundus images with labels indicating DR severity. Images were enhanced and pre-processed, and extracting features based on deep learning was done using pre-trained convolutional neural networks (CNNs). In particular, the models that were used to extract the features included ResNet50, InceptionV3, DenseNet121, DenseNet169, VGG16, Xception, and a custom Deep Neural Network (DNN). Moreover, ensemble methods were designed through integrating CNN-based feature extractors with machine learning classifiers such as Random Forest, Support Vector Machine (SVM), XGBoost and LightGBM to improve the accuracy of classification and generalization. The results of the experiments showed that among the autonomous deep learning models, Xception achieved the highest performance, with a testing accuracy of 95%. EfficientNetB3 combined with XGBoost achieved the highest testing accuracy (86.83%). This study highlights the possibilities of deploying deep learning and ensemble AIs into DR screening devices, especially in telemedicine systems and mobile diagnostic systems. Such integration could significantly improve early detection rates, reduce the burden on healthcare providers, and make diabetic eye care more accessible in underserved regions.
2022-07-01
to
2023-06-30
Research
Smart Assistance System for the Blind and Visually Impaired Person Using Image Processing and Machine Learning Approach
Abstract: According to WHO (World Health Organization) statistics, out of 8 billion people, around 284 million live with visual impairment, 39 million are blind, and 245 million have moderate to severe vision impairment worldwide. Visually impaired people face many difficulties in safe and self-supported movement. The proposed system will help such people to guide through their path, which does face recognition, object detection, and object recognition using image processing and machine learning techniques. Image Processing based solutions appear to be the most favorable options among the different utilized technologies due to their accessibility and affordability. It also provides the distance of that object from the camera. All the detected persons and objects are speech-processed using text-to-speech API and are spoken out to the user for assistance. This work tries to convert the visual world into the audio world with the possibility of informing blind people about objects and their spatial locations. This work intends to assist visually impaired people in managing day-to-day actions and navigating their surroundings. The results obtained from the proposed application boost the confidence level and comfort level of the blind; in this way, they become more independent.
Last Updated: 2026-06-28 10:58:28
No Project Found
Last Updated: 2026-06-28 10:58:28
Introduction to Programming Environment, Computer Programming with C, Object Oriented Programming Language, Database Management Systems, Microprocessor and Interfacing, Assembly Language, Digital Signal Processing, Internet and Web Programming, System Analysis and Design, Software Engineering, Digital Image Processing.
No Course Materials Found....
Honors - 2019-20
Nusrat Mahee
Thesis Title: Identification of Diabetic Retinopathy Using Deep Learning and Ensemble Model Approach
Overview: One of the most prevalent complications of diabetes and a leading cause of preventable blindness in the world is DR. This paper examines the history of the DR detection methods, the transition of traditional image-processing algorithms to the state-of-the-art deep learning and ensemble models. This study employed MESSIDOR dataset; it includes retinal fundus images with labels indicating DR severity. Images were enhanced and pre-processed, and extracting features based on deep learning was done using pre-trained convolutional neural networks (CNNs). In particular, the models that were used to extract the features included ResNet50, InceptionV3, DenseNet121, DenseNet169, VGG16, Xception, and a custom Deep Neural Network (DNN). Moreover, ensemble methods were designed through integrating CNN-based feature extractors with machine learning classifiers such as Random Forest, Support Vector Machine (SVM), XGBoost and LightGBM to improve the accuracy of classification and generalization. The results of the experiments showed that among the autonomous deep learning models, Xception achieved the highest performance, with a testing accuracy of 95%. EfficientNetB3 combined with XGBoost achieved the highest testing accuracy (86.83%). This study highlights the possibilities of deploying deep learning and ensemble AIs into DR screening devices, especially in telemedicine systems and mobile diagnostic systems. Such integration could significantly improve early detection rates, reduce the burden on healthcare providers, and make diabetic eye care more accessible in underserved regions.
Honors - 2019-20
Mrittika Das
Thesis Title: Deep Learning-Based Ensemble Framework for Automated Pneumonia Detection Using Chest X-ray Imaging
Overview: Pneumonia remains one of the leading causes of illness and death in children worldwide, with the burden particularly high in low-resource regions where radiological expertise is limited. Chest X-ray imaging is the primary diagnostic tool, but its interpretation is often subjective, inconsistent, and dependent on specialist availability. Deep learning with convolutional neural networks (CNNs) has shown promise in automating diagnosis, yet most existing studies rely on single-model frameworks that face challenges of generalizability, robustness, and clinical applicability. This study develops and evaluates two ensemble deep learning models—DenseNet121–ResNet50 and DenseNet121–EfficientNetB3—designed to improve diagnostic accuracy and reliability for pediatric pneumonia detection. A clinically motivated preprocessing pipeline, combining Contrast Limited Adaptive Histogram Equalization (CLAHE) with extensive data augmentation, was employed to enhance image quality and address variability typical of real-world X-rays. Unlike prior works that either depend solely on single CNNs or involve computationally heavy ensembles, our approach emphasizes a balanced design that is both accurate and practical for deployment in resource-constrained healthcare settings. Experimental results on a benchmark chest X-ray dataset demonstrate that the DenseNet121–EfficientNetB3 ensemble achieved 96.06% accuracy and an AUC of 0.9910, outperforming DenseNet121–ResNet50 (95.70% accuracy, 0.9898 AUC). These findings highlight the potential of ensemble CNNs, when paired with clinically relevant preprocessing, to provide dependable diagnostic support in environments where radiological expertise is scarce.
Honors - 2019-20
Hemeka Rashid Hafsa
Thesis Title: A Hybrid Deep Learning Framework: CNN–BiLSTM–Attention Model with SMOTE for Depression Detection from Social Media Posts
Overview: Depression, a ubiquitous mental illness, creates major difficulties in society, as it affects both an individual and society. Barriers to early detection include stigma, underreporting, and limited access to care. The first major advantage of social media as a medium is that it provides textual information that can be analyzed to provide evidence of depression, showing how the users feel. To identify depressed messages in social media, this article proposes a deep learning architecture. As a proxy, the engagement ratings were first assigned to a raw, dirty, and biased dataset; posts that contained a high number of likes were designated as possibly depressed, and those with a low number of likes were designated as nondepressive. Standard NLP techniques, i.e., tokenization, lemmatization, and stopword deletion, were used to clean up the text data and make it structured. To obtain both contextual and local information, a hybrid deep learning model was created that incorporated CNN, BiLSTM, and an attention layer. The semantic representation used GloVe embeddings in 100 dimensions that were pre-trained. The proposed model showed a good performance with a macro F1-score of 0.97 and an accuracy of 97.26%. These findings show how well deep learning works to detect gloomy content in words posted on social media.
Honors - 2018-19
Sourav Debnath
Thesis Title: Improving Students' Residence Hall Security with Machine Learning Techniques
Overview: Ensuring student safety in campus dormitories is vital for modern educational institutions. Our innovative Student Residence Hall Security System integrates advanced technologies for comprehensive threat detection. The Facial Recognition module combines Multi-task Cascaded Convolutional Networks (MTCNN) for precise facial feature detection and the FaceNet model for reliable recognition of authorized individuals. This ensures secure access control, allowing only verified residents and staff into dormitories. Additionally, our system uses the YOLOv8 (You Only Look Once) model for Harmful Object Detection, identifying and flagging dangerous items such as weapons or unauthorized electronics. This proactive feature prevents potential security breaches. The Gathering Detection module, powered by the MobileNet Single Shot Detector (SSD) model, monitors crowd behavior, detecting gatherings or unusual activities around residence halls in real time. This supports efficient crowd management and timely responses to potential security threats. The system's adaptability allows deployment across various campus sizes, enhancing security for students, staff, and faculty. Unlike previous research that focused solely on face recognition and detection, our research extends to high-precision identification of gatherings and harmful objects, creating a holistic and proactive security solution.
Honors - 2018-19
Katha Saha
Thesis Title: A Smart Shopping Assistant at Super Shop for Visually Challenged Persons Utilizing Machine Learning Approach
Overview: This research develops an Assistive System to help visually impaired individuals shop in super shops. The system combines machine learning, real-time image processing, and text-to-speech technologies to make shopping easier. A custom dataset with 30 categories of essential food and wearable items is used. The YOLOv8 pre-trained model, known for its strong performance, handles object detection and recognition, achieving 82.7% accuracy. When an object is detected, Google's Text- To-Speech (GTTS) technology converts visual information into spoken words, giving immediate feedback to the user. The research focuses on data collection for training, real-time object detection, and converting recognized items into speech. By addressing the challenges faced by visually impaired individuals, this study shows how image processing and machine learning can be used in assistive technologies. The system is designed to improve the quality of life for visually impaired people by making shopping more independent and confident. It offers a practical solution to help them navigate and shop in super shops, fulfilling an important need in their daily lives.
Honors - 2017-18
Umme Kulsum Shrabony
Thesis Title: Early Identification of Diabetic Retinopathy Using Deep Learning Model
Overview: Enter Project/Thesis OverviewDiabetes mellitus and its symptoms are causing severe and even life-threatening difficulties in the lives of people all over the world. A side effect of diabetes mellitus called diabetic retinopathy causes an eye condition that can impair vision. The important aspect of treatment is that if it is detected early, then disease progression can be slowed by controlling modifiable risk factors (such as blood glucose and blood pressure), and laser therapy or treatment with vascular endothelial growth factor inhibitors can be used to reduce the risk of visual loss. Early identification of diabetic retinopathy is aided by retinal screening. Recently, deep learning methods are becoming massively widely used to facilitate automatic retinal screening in the retina's fundus. Deep learning methods are more effective in the early detection of deep retinal retinopathy. To evaluate the effectiveness of deep learning approaches for detecting diabetic retinopathy, we conducted a systematic review, including a meta-analysis of distinct deep learning methods and several types of classifiers from some recent pertinent research work. This survey will be useful for technical people and academics who wish to strive to improve the diagnostics of a system that will be more powerful in practice.
- Institutional Email: sabbirejaz.ice@nstu.edu.bd
- Personal Email: cse.sabbir7@gmail.com
- Mobile number: 01725990631
- Emergency Contact: N/A
- PABX:N/A
- Website:N/A
Department
Information and Communication Engineering
Noakhali Science and Technology University