Towards Data-Eficient and Interpretable Computer Vision: Advances in Few-Shot Learning and Explainable AI.
On August 10, 2025, the Department of CSE at Green University of Bangladesh hosted a seminar titled “Towards Data-Efficient and Interpretable Computer Vision: Advances in Few-Shot Learning and Explainable AI.” Organized by IEEE Computer Society GUB SBC, co-organized by IEEE GUB SB, and in association with the AI & ML Research Cell, with technical support from IEEE CS Bangladesh Chapter, the event began at 10:30 a.m. with a warm welcome speech.
Following the welcome address, our honorable Chairperson, Professor Dr. Md. Ahsan Habib, SMIEEE, delivered a brief speech emphasizing the significance of academic events and acknowledging IEEE’s contribution in organizing the seminar. He highlighted the value of knowledge-sharing platforms in fostering students’ academic and professional growth.
The keynote speech was then delivered by the Vice-Chancellor of Green University of Bangladesh, Professor Dr. Mohammad Shorif Uddin, SMIEEE, who also served as the Chief Guest. He began by discussing Artificial Intelligence as a pivotal force shaping the Fifth Industrial Revolution, providing insights into various image processing systems and their operational mechanisms. He elaborated on different learning paradigms, including Few-Shot Learning (FSL), and outlined the seminar’s content.
Professor Uddin explored computer vision, explaining challenges such as viewpoint variation, occlusion, scaling, deformation, intra-class variation, local ambiguity, and background complexity. He traced the historical evolution of computer vision and emphasized how machine learning can enhance work efficiency. Encouraging students to engage with academic literature, he discussed supervised, unsupervised, semi-supervised, and reinforcement learning, supported by visual graphs and frameworks.
The discussion extended to deep learning, conventional machine learning algorithms, transfer learning, Generative Adversarial Networks (GANs), and federated learning, highlighting research opportunities in these areas. He stressed data efficiency and interpretability, addressing limitations of current approaches and strategies like data augmentation and synthetic data generation through GANs.
Professor Uddin provided a detailed explanation of Few-Shot Learning, covering meta-learning, metric-based learning, optimization-based approaches, and generative methods with illustrative diagrams. He then introduced Explainable AI (XAI), highlighting tools such as SHAP and LIME for transparency and accountability in AI systems. He concluded by emphasizing the synergy between FSL and XAI in developing AI that is both powerful and interpretable.
Following the keynote, Professor Dr. Md. Ahsan Habib delivered the closing speech, reinforcing the importance of research, advising students to develop independence for securing internships, and encouraging proactive career development. He also discussed improvements to the university’s multipurpose room facilities and thanked all participants.The seminar concluded with an award ceremony and photo session, recognizing students who completed IDP-1 and IDP-2 for their outstanding projects. The event officially ended at 1:00 p.m., leaving participants enriched with insights into the evolving landscape of AI, Few-Shot Learning, and Explainable AI.
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
- Green University of Bangladesh
- Purbachal American City, Dhaka
- Bangladesh
- Building: Multi-Purpose hall
- Starts 31 July 2025 12:00 PM UTC
- Ends 08 August 2025 06:00 PM UTC
- No Admission Charge
Speakers
Shorif Uddin
Towards Data-Eficient and Interpretable Computer Vision: Advances in Few-Shot Learning and Explainable AI
The keynote speech was delivered by the Vice-Chancellor of Green University of Bangladesh, Professor Dr. Mohammad Shorif Uddin, SMIEEE, who also served as the Chief Guest for the event. He began by discussing Artificial Intelligence as a pivotal force shaping the Fifth Industrial Revolution, highlighting various image processing systems and their operational mechanisms. He elaborated on different learning paradigms, including Few-Shot Learning (FSL), and provided an outline of the seminar content.
Professor Uddin explored computer vision in depth, explaining challenges such as viewpoint variation, occlusion, scaling, deformation, intra-class variation, local ambiguity, and background complexity. He traced the historical evolution of computer vision and emphasized how machine learning can significantly enhance work efficiency. Encouraging students to read academic literature and stay updated with technological advancements, he discussed supervised, unsupervised, semi-supervised, and reinforcement learning, supported by visual graphs and frameworks.
The discussion further extended to deep learning, conventional machine learning algorithms, transfer learning, Generative Adversarial Networks (GANs), and federated learning, highlighting the research potential of these areas. He stressed the importance of data efficiency and interpretability, outlining limitations of current approaches and strategies such as data augmentation and synthetic data generation through GANs.
The session then focused on Few-Shot Learning, covering meta-learning, metric-based learning, optimization-based approaches, and generative methods, supported by illustrative diagrams. Professor Uddin concluded with Explainable AI (XAI), highlighting SHAP and LIME as tools for transparency and accountability in AI systems, and emphasized the synergy between FSL and XAI in creating AI that is both powerful and interpretable.
The seminar successfully bridged theory and practice, offering insights into the latest advancements in AI-powered computer vision. It served as a platform for knowledge exchange, inspiring curiosity and motivating students toward research and innovation. By uniting experts, faculty, and aspiring engineers, the event significantly contributed to fostering a research-driven and forward-thinking academic environment at Green University of Bangladesh.