Muzammil Behzad
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
Muzammil Behzad has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
- Guoying Zhao
University of Oulu, Oulu, Finland
5 shared publications - Xiaobai Li
Zhejiang University, Hangzhou, China and University of Oulu, Oulu, Finland
5 shared publications - Nhat Vo
Center for Machine Vision and Signal Analysis (CMVS) University of Oulu Finland
3 shared publications Abdullah Naeem Malik
1 shared publication- Afrah Shaahid
College of Computer Science and Information Technology, Imam Abdulrahman Bin Faisal University, P.O. Box 1982, Dammam 31441
1 shared publication Alaa Dalaq
King Fahd University of Petroleum and Minerals, Saudi Arabia
1 shared publication
Latest Publications
- 8 October 2026
- Published by Elsevier in Neural Networks
- Vol. 206 (Pt A) , 109721
SegVLM: Deformable attentive visual enhancement for referring image segmentation using vision-language model
- 8 October 2026
- Published by Elsevier in Neural Networks
- Vol. 206 (Pt A) , 109721
SegVLM: Deformable attentive visual enhancement for referring image segmentation using vision-language model
- Top Cited Papers
- 28 March 2026
- Published by Elsevier in Pattern Recognition
- Vol. 179
UDAN-CLIP: Diffusion attention network with contrastive language-image joint learning for underwater image enhancement
- 28 March 2026
- Published by Elsevier in Pattern Recognition
- Vol. 179
Top Cited PapersUDAN-CLIP: Diffusion attention network with contrastive language-image joint learning for underwater image enhancement
- 25 March 2026
- Published by Elsevier in Pattern Recognition
- Vol. 179
KRCapVLM: Beam-guided knowledge replay for knowledge-rich image captioning using vision-language model
- 25 March 2026
- Published by Elsevier in Pattern Recognition
- Vol. 179
KRCapVLM: Beam-guided knowledge replay for knowledge-rich image captioning using vision-language model
- 19 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-10
MultiviewVLM: Unsupervised Multiview Contrastive Language-Image Joint Learning with Pseudo-Labeled Prompts Via Vision-Language Model for 3D/4D Facial Expression Recognition
In this paper, we introduce MultiviewVLM, a vision language model designed for unsupervised contrastive multiview representation learning of facial emotions from 3D/4D data. Our architecture integrates pseudo-labels derived from generated textual prompts to guide implicit alignment of emotional semantics. To capture shared information across multi-views, we propose a joint embedding space that aligns multiview representations without requiring explicit supervision. We further enhance the discriminability of our model through a novel multiview contrastive learning strategy that leverages stable positive-negative pair sampling. A gradient-friendly loss function is introduced to promote smoother and more stable convergence, and the model is optimized for distributed training to ensure scalability. Extensive experiments demonstrate that MultiviewVLM outperforms existing state-of-the-art methods and can be easily adapted to various real-world applications with minimal modifications.
- 19 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-10
MultiviewVLM: Unsupervised Multiview Contrastive Language-Image Joint Learning with Pseudo-Labeled Prompts Via Vision-Language Model for 3D/4D Facial Expression Recognition
In this paper, we introduce MultiviewVLM, a vision language model designed for unsupervised contrastive multiview representation learning of facial emotions from 3D/4D data. Our architecture integrates pseudo-labels derived from generated textual prompts to guide implicit alignment of emotional semantics. To capture shared information across multi-views, we propose a joint embedding space that aligns multiview representations without requiring explicit supervision. We further enhance the discriminability of our model through a novel multiview contrastive learning strategy that leverages stable positive-negative pair sampling. A gradient-friendly loss function is introduced to promote smoother and more stable convergence, and the model is optimized for distributed training to ensure scalability. Extensive experiments demonstrate that MultiviewVLM outperforms existing state-of-the-art methods and can be easily adapted to various real-world applications with minimal modifications.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 144420-144433
Attack-Aware Deepfake Detection Under Counter-Forensic Manipulations
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 144420-144433
Open AccessAttack-Aware Deepfake Detection Under Counter-Forensic Manipulations
- 27 September 2025
- Published by Elsevier in Neurocomputing
- Vol. 657
FACET–VLM: Facial emotion learning with text-guided multiview fusion via vision-language model for 3D/4D facial expression recognition
- 27 September 2025
- Published by Elsevier in Neurocomputing
- Vol. 657
FACET–VLM: Facial emotion learning with text-guided multiview fusion via vision-language model for 3D/4D facial expression recognition
- Open Access
- 12 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 143831-143842
SMILE-VLM: Self-Supervised Multi-View Representation Learning Using Vision-Language Model for 3D/4D Facial Expression Recognition
Facial expression recognition (FER) is a fundamental task in affective computing with applications in human-computer interaction, mental health analysis, and behavioral understanding. In this paper, we propose SMILE-VLM, a self-supervised vision-language model for 3D/4D FER that unifies multiview visual representation learning with natural language supervision. SMILE-VLM learns robust, semantically aligned, and view-invariant embeddings by proposing three core components: multiview decorrelation via a Barlow Twins-style loss, vision-language contrastive alignment, and cross-modal redundancy minimization. Our framework achieves the state-of-the-art performance on multiple benchmarks. We further extend SMILE-VLM to the task of 4D micro-expression recognition (MER) to recognize the subtle affective cues. The extensive results demonstrate that SMILE-VLM not only surpasses existing unsupervised methods but also matches or exceeds supervised baselines, offering a scalable and annotation-efficient solution for expressive facial behavior understanding.
- 12 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 143831-143842
Open AccessSMILE-VLM: Self-Supervised Multi-View Representation Learning Using Vision-Language Model for 3D/4D Facial Expression Recognition
Facial expression recognition (FER) is a fundamental task in affective computing with applications in human-computer interaction, mental health analysis, and behavioral understanding. In this paper, we propose SMILE-VLM, a self-supervised vision-language model for 3D/4D FER that unifies multiview visual representation learning with natural language supervision. SMILE-VLM learns robust, semantically aligned, and view-invariant embeddings by proposing three core components: multiview decorrelation via a Barlow Twins-style loss, vision-language contrastive alignment, and cross-modal redundancy minimization. Our framework achieves the state-of-the-art performance on multiple benchmarks. We further extend SMILE-VLM to the task of 4D micro-expression recognition (MER) to recognize the subtle affective cues. The extensive results demonstrate that SMILE-VLM not only surpasses existing unsupervised methods but also matches or exceeds supervised baselines, offering a scalable and annotation-efficient solution for expressive facial behavior understanding.
- Open Access
- 14 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 14 (4) , 3031-3047
4DME: A Spontaneous 4D Micro-Expression Dataset With Multimodalities
Micro-expressions (ME) are a special form of facial expressions which may occur when people try to hide their true feelings for some reasons. MEs are important clues to reveal people’s true feelings, but are difficult or impossible to be captured by ordinary persons with naked-eyes as they are very short and subtle. It is expected that robust computer vision methods can be developed to automatically analyze MEs which requires lots of ME data. The current ME datasets are insufficient, and mostly contain only one single form of 2D color videos. Researches on 4D data of ordinary facial expressions have prospered, but so far no 4D data is available in ME study. In the current study, we introduce the 4DME dataset: a new spontaneous ME dataset which includes 4D data along with three other video modalities. Both micro- and macro-expression clips are labeled out in 4DME, and 22 AU labels and five categories of emotion labels are annotated. Experiments are carried out using three 2D-based methods and one 4D-based method to provide baseline results. The results indicate that the 4D data can potentially benefit ME recognition. The 4DME dataset could be used for developing 4D-based approaches, or exploring fusion of multiple video sources (e.g., texture and depth) for the task of ME analysis in future. Besides, we also emphasize the importance of forming a clear and unified criteria of ME annotation for future ME data collection studies. Several key questions related with ME annotation are listed and discussed in depth, especially about the relationship between AUs and ME emotion categories. A preliminary AU-Emo mapping table is proposed with justified explanations and supportive experimental results. Several unsolved issues are also summarized for future work.
- 14 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. 14 (4) , 3031-3047
Open Access4DME: A Spontaneous 4D Micro-Expression Dataset With Multimodalities
Micro-expressions (ME) are a special form of facial expressions which may occur when people try to hide their true feelings for some reasons. MEs are important clues to reveal people’s true feelings, but are difficult or impossible to be captured by ordinary persons with naked-eyes as they are very short and subtle. It is expected that robust computer vision methods can be developed to automatically analyze MEs which requires lots of ME data. The current ME datasets are insufficient, and mostly contain only one single form of 2D color videos. Researches on 4D data of ordinary facial expressions have prospered, but so far no 4D data is available in ME study. In the current study, we introduce the 4DME dataset: a new spontaneous ME dataset which includes 4D data along with three other video modalities. Both micro- and macro-expression clips are labeled out in 4DME, and 22 AU labels and five categories of emotion labels are annotated. Experiments are carried out using three 2D-based methods and one 4D-based method to provide baseline results. The results indicate that the 4D data can potentially benefit ME recognition. The 4DME dataset could be used for developing 4D-based approaches, or exploring fusion of multiple video sources (e.g., texture and depth) for the task of ME analysis in future. Besides, we also emphasize the importance of forming a clear and unified criteria of ME annotation for future ME data collection studies. Several key questions related with ME annotation are listed and discussed in depth, especially about the relationship between AUs and ME emotion categories. A preliminary AU-Emo mapping table is proposed with justified explanations and supportive experimental results. Several unsolved issues are also summarized for future work.
- Open Access
- 26 January 2022
- Published by MDPI AG in Preprints.org
Technology-Embedded Hybrid Learning
With the rapid surge in technological advancements, an equal amount of investment in technology-embedded teaching has become vital to pace up with the ongoing educational needs. Distance education has evolved from the era of postal services to the use of ICT tools in current times. With the aid of globally updated content across the board, technology usage ensures all students receive equal attention without any discrimination. Importantly, web-based teaching allows all kind of students to learn at their own pace, without the fear of being judged, including professionals who can learn remotely without disturbing their job schedules. Having web-based content allows low-cost and robust implementation of the content upgradation. An improved, yet effective, version of the education using such tools is Hybrid Learning (HL). This learning mode aims to provide luxurious reinforcement to its legitimate candidates while maintaining the quality standards of various elements. Incorporated with both traditional and distance learning methods, along with exploiting social media tools for increased comfort level and peer-to-peer collaboration, HL ultimately facilitates the end user and educational setup. The structure of such a hybrid model is realized by delivering the study material via a learning management system (LMS) designed in compliance with quality standards, which is one of the fundamental tackling techniques for controlling quality constraints. In this paper, we present the recently piloted project by COMSATS University Islamabad (previously known as COMSATS Institute of Information Technology) which is driven by technology-embedded teaching model. This model is an amalgam of the traditional class room model with the aid of state-of-the-art online learning technologies. The students are enrolled as full-time students, with all the courses in traditional classroom mode, except one course offered as hybrid course. This globally adapted model helps the students to benefit from both face-to-face learning as well as gaining hands-on experience on technology-enriched education model providing flexibility of timings, learning pace, and boundaries. Our HL model is equipped with two major synchronous and asynchronous blocks. The synchronous block delivers real-time live interaction scenarios using discussion boards, thereby providing a face-to-face environment. Interactions via social network has witnessed equally surging improvement in the output performance. The asynchronous block refers to the lecture videos, slides and handouts, prepared by imminent professors, available 24/7 for students. To ensure quality output, our HL model follows the course learning outcomes (CLOs), and program learning outcomes (PLOs) as per international standards. As a proof of concept, we have deployed a mechanism at the end of each semester to verify the effectiveness of our model. This mechanism fundamentally surveys the satisfaction levels of all the students enrolled in the HL courses. With the surveys already conducted, a significant level of satisfaction has been noted. Extensive results from these surveys are presented in the paper to further validate the efficiency and robustness of our proposed HL model.
- 26 January 2022
- Published by MDPI AG in Preprints.org
Open AccessTechnology-Embedded Hybrid Learning
With the rapid surge in technological advancements, an equal amount of investment in technology-embedded teaching has become vital to pace up with the ongoing educational needs. Distance education has evolved from the era of postal services to the use of ICT tools in current times. With the aid of globally updated content across the board, technology usage ensures all students receive equal attention without any discrimination. Importantly, web-based teaching allows all kind of students to learn at their own pace, without the fear of being judged, including professionals who can learn remotely without disturbing their job schedules. Having web-based content allows low-cost and robust implementation of the content upgradation. An improved, yet effective, version of the education using such tools is Hybrid Learning (HL). This learning mode aims to provide luxurious reinforcement to its legitimate candidates while maintaining the quality standards of various elements. Incorporated with both traditional and distance learning methods, along with exploiting social media tools for increased comfort level and peer-to-peer collaboration, HL ultimately facilitates the end user and educational setup. The structure of such a hybrid model is realized by delivering the study material via a learning management system (LMS) designed in compliance with quality standards, which is one of the fundamental tackling techniques for controlling quality constraints. In this paper, we present the recently piloted project by COMSATS University Islamabad (previously known as COMSATS Institute of Information Technology) which is driven by technology-embedded teaching model. This model is an amalgam of the traditional class room model with the aid of state-of-the-art online learning technologies. The students are enrolled as full-time students, with all the courses in traditional classroom mode, except one course offered as hybrid course. This globally adapted model helps the students to benefit from both face-to-face learning as well as gaining hands-on experience on technology-enriched education model providing flexibility of timings, learning pace, and boundaries. Our HL model is equipped with two major synchronous and asynchronous blocks. The synchronous block delivers real-time live interaction scenarios using discussion boards, thereby providing a face-to-face environment. Interactions via social network has witnessed equally surging improvement in the output performance. The asynchronous block refers to the lecture videos, slides and handouts, prepared by imminent professors, available 24/7 for students. To ensure quality output, our HL model follows the course learning outcomes (CLOs), and program learning outcomes (PLOs) as per international standards. As a proof of concept, we have deployed a mechanism at the end of each semester to verify the effectiveness of our model. This mechanism fundamentally surveys the satisfaction levels of all the students enrolled in the HL courses. With the surveys already conducted, a significant level of satisfaction has been noted. Extensive results from these surveys are presented in the paper to further validate the efficiency and robustness of our proposed HL model.