Our project has been officially accepted and funded under the TÜBİTAK 3501 (National Young Researchers Career Development Program).
Project Title: Detection of Partially Inserted Segments and Completely Fake Content Through Temporal-Spatial Analysis with an Image and Video-Based Multimodal Approach
Project Duration: 24 Months (2 Years)
Project Team
- Principal Investigator (Executor): Assoc. Prof. Dr. İbrahim Delibaşoğlu (Sakarya University, Software Engineering)
- Advisor: Prof. Dr. Ahmet Özmen (Sakarya University, Software Engineering)
- Researcher: İrfan Kösesoy (Kocaeli University, Software Engineering)
Project Overview and Scope
Deepfake manipulations are evolving rapidly, moving beyond full-video fakes to highly sophisticated partial fakes—such as injecting fake content into specific segments (splicing) or removing existing objects (inpainting).
Over the next 24 months, our team will work on developing a multimodal and modular deepfake detection platform. By integrating temporal inconsistency analysis, image-level manipulation detection, and audio-visual desynchronization into a single hierarchical decision mechanism, this platform will push the boundaries of frame-level localization in complex video structures. The project will also deliver a brand-new, unique dataset to the scientific community to address current benchmarks' limitations.
Empowering Next-Gen Researchers
As part of our commitment to academic growth and hands-on research, this 2-year project will fully fund and support:
- 2 Graduate Scholarship Recipients (Ph.D. / M.Sc. level)
- 3 Bachelor's Scholarship Recipients (Undergraduate level)
These students will gain direct, real-world experience in cutting-edge deep learning, computer vision, and multimodal data analysis within ML LAB (Machine Learning & Intelligent Systems Laboratory).
