2025 International Conference on Image Processing and Deep Learning(IPDL) has successfully concluded!

From April 11 toApril 13, 2025, the 2025 International Conference on Image Processing and Deep Learning(IPDL) hosted by Southwest University , were successfully held in Chengdu, China.

The event was marked by a highly engaging atmosphere, with scholars and experts actively discussing the latest research and applications in the fields of Image Processing and Deep Learning. It was a fruitful exchange of ideas and insights!

The conference featured outstanding keynote presentations by renowned experts, including Professor Jianping Gou, IEEE Senior Member,Southwest University,China; Professor Yulin Wang, IEEE Senior Member, Wuhan University, China; Professor Lu Leng, IEEE Member,Nanchang Hangkong University,China and associate Professor Zhi Wang from Southwest University,China.

ProfessorJianping Gou summarizes knowledge distillation as a simple and effective technique for compressing large models in deep learning. It provides an overview of large models and model compression, followed by a detailed discussion of the theory, classical algorithms, and application scenarios of model distillation, and concludes with insights into its future development.

Lu Leng mentioned that reusable biometric templates offer a robust solution to the security and privacy challenges posed by traditional biometric data storage methods. By employing irreversible transformations and aligning with global standards, they provide a -centric approach that enhances trust and compliance. However, further exploration into their technical processes, performance trade-offs, and mechanisms for handling data changes is necessary to fully appreciate their significance and implementation across various industries.

In this talk, a general low rank matrix learning (LRML) model is introduced by Prof. Zhi Wang. The model incorporates a novel nonconvex regularizer that considers the varying contributions of different rank components and adaptively truncates sequential singular values. Additionally, an optimization method with high computational efficiency and convergence guarantees is developed to solve the LRML model. By leveraging the Kurdyka-Łojasiewicz (KŁ) inequality, the local and global convergence properties of the method are established. The proposed approach is demonstrated to be applicable to various low rank tasks, including matrix completion and subspace clustering.

Prof. Yulin Wang cover some of the earliest and simplest forms of data hiding in digital multimedia and then move to some of the lasted innovations in order to provide insight into these questions. Some of the research branches, called reversible data hiding, is also depicted.

The successful hosting of this conference provides a high-end academic exchange platform for professionals in the fields of Image Processing and Deep Learning. In the future, we look forward to more international academic conferences like IPDL 2025, which will continue to promote academic exchanges and cooperation worldwide, and jointly promote technological innovation and development in these fields. I believe that with the joint efforts of all sectors, the future of Image Processing and Deep Learning will be full of infinite possibilities, making greater contributions to the progress of human society.

Thank you to all the speakers, attendees, and staff who participated in this conference. Although IPDL 2025 has come to an end, our exploration and research in the fields of Image Processing and Deep Learning and Computational Intelligence will never stop. Looking forward to our next gathering!