Special Issue on
Generative AI and Large Vision-Language Models for Biometrics
Submission Deadline: 31 May 2025
Targeted Publication: Q1 2026
Paper submission: https://ieee.atyponrex.com/journal/tbiom
*********************************************************************************************
*** Motivation ***
In the rapidly advancing field of artificial intelligence, generative AI
and large-scale vision-language models are becoming key areas of
interest, revolutionizing numerous research fields, including natural
language processing and computer vision. Generative AI models are
designed and trained to approximate the underlying distribution of a
dataset, enabling the generation of new samples that reflect the
patterns and regularities within the training data. Among the various
types of generative models, such as Generative Adversarial Networks
(GANs), Variational Autoencoders (VAEs), flow-based, autoregressive, and
diffusion models, GANs and diffusion models have gained significant
attention and are widely applied to tasks such as image synthesis, image
manipulation, text generation, and speech synthesis. These models have
shown remarkable success in modeling and interpreting the probability
distributions of real-world data. Vision-language models, on the other
hand, integrate visual and textual data, learning to associate these
modalities to enhance understanding and enable multimodal
reasoning-based applications.
The advancements in generative AI and vision-language models (LVMs) are
also making a significant impact on biometrics, offering new
possibilities for addressing longstanding challenges. Generative AI,
with its ability to synthesize highly realistic data, has the potential
to address privacy concerns related to collecting, sharing, and using
sensitive biometric data. This synthetic data can also be used to
increase diversity and variation in training datasets through
augmentation, thus improving model generalizability and reducing
potential bias induced by imbalanced training data. At the same time,
large vision-language models offer the capability to process and
understand multimodal information by combining visual features with
contextual data, such as semantic insights from natural language.
Furthermore, large-scale vision-language models can be optimized for
downstream tasks, such as template extraction, using zero or few-shot
learning approaches, making them highly versatile for biometric
applications.
Although generative AI and vision-language models offer a rich set of
tools that can be utilized to address challenges in biometrics, the
misuse of these technologies presents a threat to the field. Generative
AI models have the ability to incorporate conditions in the generation
process to take control over the generated samples. This enables a wide
range of applications such as image-to-image translation, text-to-image
synthesis, and style transfer. However, this capability also allows for
creating deepfake attacks, e.g., images, videos, and audio that are
indistinguishable or nearly indistinguishable from real content. The
increased realism and widespread public accessibility of generative AI
have raised concerns about the potential misuse of this technology for
malicious purposes. This highlights the need for solutions to detect
generated AI content and mitigate the potential misuse of generative AI
models.
The proposed TBIOM special issue will provide a platform to discuss the
latest advancements and technical achievements related to Generative AI
and Large vision-language models when applied to problems in biometrics.
The topics of interest of the special issue include, but are not limited to:
+ Novel generative AI models for responsible synthesis of biometric data
+ Novel generative models for conditional data synthesis
+ Biometrics interpretability and explainability through large
language-vision models
+ Few-shot learning from large language-vision models
+ Generative AI and LVMs for detecting attacks on biometrics systems
+ Generative AI-based image restoration
+ Information leakage of synthetic data
+ Data factories and label generation for biometric models
+ Quality assessment of AI generated data
+ Synthetic data for data augmentation
+ Detection of generated AI contents
+ Bias mitigation using synthetic data
+ LLMs and VLMs for biometrics
+ Watermarking AI generated content
+ New synthetic datasets and performance benchmarks
+ Security and privacy issues regarding the use of generative AI methods
for biometrics
+ Ethical considerations regarding the use of generative AI methods for
biometrics
+ Parameter efficient fine-tuning of VLMs for biometrics applications
*** Important Dates ***
Submission deadline: 31 May 2025
First round of reviews completed (first decision): August 2025
Second round of reviews completed October 2025
Final papers due December 2025
Publication date: Q1 2026
*** Paper Submission ***
Papers should be submitted through the TBIOM submission portal before
the deadline using the TBIOM journal templates:
https://ieee.atyponrex.com/journal/tbiom and selecting the article type:
“Generative AI and Large Vision-Language Models for Biometrics”.
*** Guest Editors: ***
+ Fadi Boutros, Fraunhofer IGD, Germany
+ Hu Han, Institute of Computing Technology, Chinese Academy of Sciences
(CAS), China
+ Tempestt Neal, University of South Florida, United States
+ Vishal M. Patel, Johns Hopkins University, United States
+ Vitomir Štruc, University of Ljubljana, Slovenia
+ Yunhong Wang, Beihang University, China