eXeLMM

Despite the remarkable advancements of deep learning methodologies over the past decade, in-depth video understanding remains a challenge. eXeLMM will address this challenge by harnessing the power of Large Language Models (LLMs) / Large Multimodal Models (LMMs). Powerful such open models will be selected, adapted and employed for achieving in-depth video understanding. As part of this work, eXeLMM will make considerable advancements in relation to critical issues that arise when attempting to exploit LLMs/LMMs in video tasks, such as computational efficiency and explainability.

• Advance video event recognition / video understanding, by adapting and employing powerful LLMs and LMMs.
• Reduce the computational cost of LLM/LMM-based video understanding, e.g., by introducing knowledge distillation approaches.
• Improve the trustworthiness of video understanding, by introducing AI explainability techniques.
• Exploit multimodality towards better video understanding.

Dr. Vasileios Mezaris, Research Director (Researcher A), eXeLMM Principal Investigator
Head of the Intelligent Digital Transformation (IDT) Laboratory of ITI-CERTH
PhD Electrical and Computer Engineering, BSc Electrical and Computer Engineering

Dr. Evlampios Apostolidis, Postdoctoral Researcher
PhD Electronic Engineering, MSc Information Systems, BSc Electrical and Computer Engineering

Andreas Goulas, PhD Candidate
BSc Electrical and Computer Engineering

Links το the project’s publications, public deliverables (upon approval) and other research materials will be published here.

eXeLMM: “Explainable and efficient large multimodal models for downstream video recognition tasks” is a 36-month research project (Oct. 2025 - Sept. 2028), funded by the Hellenic Foundation for Research and Innovation (H.F.R.I.). The project is implemented in the framework of H.F.R.I call “3rd Call for H.F.R.I.’s Research Projects to Support Faculty Members & Researchers” (H.F.R.I. Project Number: 25957).