Ashraful Alam
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ashraful Alam has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Golam Rabiul Alam
BRAC University
2 shared publications - Shakib Mahmud Dipto
Old Dominion University, Norfolk, Virginia, USA
2 shared publications - Farhan Faruk
Department of Computer Science and Engineering, BRAC University, Dhaka, Bangladesh
1 shared publication Farhan Uz Zaman
BRAC University,Department of Computer Science and Engineering,Dhaka,Bangladesh
1 shared publication- H. M. Sarwer Alam
Department of Computer Science and Engineering, BRAC University, Dhaka, Bangladesh
1 shared publication - Irfana Afifa
BRAC University
1 shared publication
Latest Publications
- 1 June 2026
- Published by Ital Publication in Emerging Science Journal
- Vol. 10 (3) , 1239-1262
LeukocyteNet: An Explainable Transfer-Transformer Fusion Learning Model for Leukocyte Classification
White Blood Cells (WBCs), or leukocytes, are essential components of the immune system that protect the body against infections and malignant disorders. Even minor fluctuations in leukocyte count can indicate serious pathological conditions, including life-threatening malignancies such as leukemia, lymphoma, and myelodysplastic syndromes. Conventional diagnosis through manual microscopic examination is time-consuming, subjective, and heavily dependent on the pathologist’s expertise. To overcome these challenges, this study introduces LeukocyteNet, a transfer–transformer fusion model designed for the automated classification of ten malignant leukocyte categories. The model integrates convolutional feature extraction from VGG19 with the Swin Transformer’s global attention mechanism, enabling robust representations of both local morphology and global spatial dependencies. The LeukocyteNet model was trained on three publicly available datasets “ALL-IDB, the American Society of Hematology Image Bank, and Tehran Taleqani Hospital” and achieved an overall accuracy of 97.34%, a macro-averaged F1-score of 0.95, and a recall of 0.93, outperforming all evaluated baseline models. Furthermore, the inclusion of explainable AI techniques Grad-CAM, LIME, and Saliency Map enhances explainability by visualizing class-specific decision regions, thereby increasing clinical transparency and reliability. These findings demonstrate that LeukocyteNet not only achieves state-of-the-art predictive performance but also provides interpretable insights critical for trustworthy medical diagnostics.
- 1 June 2026
- Published by Ital Publication in Emerging Science Journal
- Vol. 10 (3) , 1239-1262
LeukocyteNet: An Explainable Transfer-Transformer Fusion Learning Model for Leukocyte Classification
White Blood Cells (WBCs), or leukocytes, are essential components of the immune system that protect the body against infections and malignant disorders. Even minor fluctuations in leukocyte count can indicate serious pathological conditions, including life-threatening malignancies such as leukemia, lymphoma, and myelodysplastic syndromes. Conventional diagnosis through manual microscopic examination is time-consuming, subjective, and heavily dependent on the pathologist’s expertise. To overcome these challenges, this study introduces LeukocyteNet, a transfer–transformer fusion model designed for the automated classification of ten malignant leukocyte categories. The model integrates convolutional feature extraction from VGG19 with the Swin Transformer’s global attention mechanism, enabling robust representations of both local morphology and global spatial dependencies. The LeukocyteNet model was trained on three publicly available datasets “ALL-IDB, the American Society of Hematology Image Bank, and Tehran Taleqani Hospital” and achieved an overall accuracy of 97.34%, a macro-averaged F1-score of 0.95, and a recall of 0.93, outperforming all evaluated baseline models. Furthermore, the inclusion of explainable AI techniques Grad-CAM, LIME, and Saliency Map enhances explainability by visualizing class-specific decision regions, thereby increasing clinical transparency and reliability. These findings demonstrate that LeukocyteNet not only achieves state-of-the-art predictive performance but also provides interpretable insights critical for trustworthy medical diagnostics.
- Open Access
- 3 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Exploring the Impact of Display Field-of-view and Display Type on Text Readability and Visual Text Search on Large Displays in Virtual Reality: Impact of Display Field-of-view and Type on Text Readability and Search in Large VR Displays
Large curved displays enhance the viewing experience beyond flat screens by more closely aligning with the human field of vision. Despite the growing prevalence of curved displays in immersive environments, there remains a significant gap in understanding how different display field-of-views (i.e., quarter circular, semicircular, three-fourth circular, fully circular) influence user experience, particularly regarding reading comprehension and visual search performance in VR. This study examines text readability and visual text search across four curved field-of-views (90°, 180°, 270°, 360°) and their flat counterparts. Findings indicate that narrower field-of-views significantly improve reading speed over wider ones, and curved displays outperform flat ones in search accuracy. Curve90 (90°) was preferred for its lower subjective workload. Participants preferred the Curve90 display due to its ergonomic advantages and lower perceived effort, while flat displays with wider field-of-views were found to be more demanding. These results offer guidelines for optimizing reading and search tasks in virtual displays.
- 3 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessExploring the Impact of Display Field-of-view and Display Type on Text Readability and Visual Text Search on Large Displays in Virtual Reality: Impact of Display Field-of-view and Type on Text Readability and Search in Large VR Displays
Large curved displays enhance the viewing experience beyond flat screens by more closely aligning with the human field of vision. Despite the growing prevalence of curved displays in immersive environments, there remains a significant gap in understanding how different display field-of-views (i.e., quarter circular, semicircular, three-fourth circular, fully circular) influence user experience, particularly regarding reading comprehension and visual search performance in VR. This study examines text readability and visual text search across four curved field-of-views (90°, 180°, 270°, 360°) and their flat counterparts. Findings indicate that narrower field-of-views significantly improve reading speed over wider ones, and curved displays outperform flat ones in search accuracy. Curve90 (90°) was preferred for its lower subjective workload. Participants preferred the Curve90 display due to its ergonomic advantages and lower perceived effort, while flat displays with wider field-of-views were found to be more demanding. These results offer guidelines for optimizing reading and search tasks in virtual displays.
- 8 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Real world objects augmentation in virtual 3D environment RealSense SDK, Deep Learning and Game Engine
We propose and demonstrate an approach for real world objects augmentation in a virtual 3D environment using deep neural networks and game engine. The proposed system consists of three basic parts: acquisition, processing and visualization with VR. The processing unit comprises translation of 2D to 3D images of real world objects and augmentation of the real world object in a virtual environment with unity. Through deep learning and RealSense SDK, skeleton data is achieved which helps augment the real world 3D object into a virtual environment in unity. The proposed system enables real-time augmentation of the real world objects in a virtual 3D environment created in unity game engine. Therefore, the system allows users developing mixed reality applications for real life problem solving.
- 8 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Real world objects augmentation in virtual 3D environment RealSense SDK, Deep Learning and Game Engine
We propose and demonstrate an approach for real world objects augmentation in a virtual 3D environment using deep neural networks and game engine. The proposed system consists of three basic parts: acquisition, processing and visualization with VR. The processing unit comprises translation of 2D to 3D images of real world objects and augmentation of the real world object in a virtual environment with unity. Through deep learning and RealSense SDK, skeleton data is achieved which helps augment the real world 3D object into a virtual environment in unity. The proposed system enables real-time augmentation of the real world objects in a virtual 3D environment created in unity game engine. Therefore, the system allows users developing mixed reality applications for real life problem solving.
- Open Access
- 8 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 8, 1-6
Multi-modal Emotion Recognition for Determining Employee Satisfaction
Emotion recognition has been popular in the field of research for quite a while now. In this paper bi-modal emotion detection has been used to find employee satisfaction in workplaces. Interviews were taken of employees from different workplaces and were recorded. The recorded interviews were then used to detect the emotions of the employees from which their satisfaction level was derived. From the interviews, six different entities of emotion were detected, which are: Happiness, Sadness, Neutral, Disgust, Anger and Surprise. Two separate independent neural networks have been utilised. In one the facial expressions were detected and in the other sentiment analysis was done on the speech of the interviews after converting it into text. The extracted emotions were then fed into a Support Vector Machine (SVM) for determining the satisfaction of the employees. The satisfactions were categorized into five different levels which are: Highly dissatisfied, Dissatisfied, Neutral, Satisfied and Highly satisfied.
- 8 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 8, 1-6
Open AccessMulti-modal Emotion Recognition for Determining Employee Satisfaction
Emotion recognition has been popular in the field of research for quite a while now. In this paper bi-modal emotion detection has been used to find employee satisfaction in workplaces. Interviews were taken of employees from different workplaces and were recorded. The recorded interviews were then used to detect the emotions of the employees from which their satisfaction level was derived. From the interviews, six different entities of emotion were detected, which are: Happiness, Sadness, Neutral, Disgust, Anger and Surprise. Two separate independent neural networks have been utilised. In one the facial expressions were detected and in the other sentiment analysis was done on the speech of the interviews after converting it into text. The extracted emotions were then fed into a Support Vector Machine (SVM) for determining the satisfaction of the employees. The satisfactions were categorized into five different levels which are: Highly dissatisfied, Dissatisfied, Neutral, Satisfied and Highly satisfied.
- 9 October 2021
- Published by Springer Nature
- p. 380-391
Interpretable COVID-19 Classification Leveraging Ensemble Neural Network and XAI
- 9 October 2021
- Published by Springer Nature
- p. 380-391
Interpretable COVID-19 Classification Leveraging Ensemble Neural Network and XAI