Multimodal AI for Human Communication
My research focuses on multimodal AI for human communication and interaction. I am particularly interested in developing data-efficient and generalizable learning methods that integrate linguistic, visual, acoustic, and motion-based behavioral signals.
My current work explores multimodal foundation models, sign language processing, and computational modeling of human communication, with broader interests in representation learning, low-resource learning, and socially intelligent AI.
News
-
September 2026
Joined KTH Speech, Music and Hearing as a postdoc.
-
August 2026
A master's student, co-supervised with Zerrin Yumak (Utrecht University), graduated. Congrats Cem!
-
March 2026
MSCA PF 2025 – Seal of Excellence received from the European Commission.
-
February 2026
A national patent has been registered.
-
January 2026
New paper at Expert Systems with Applications: How well do vision models understand tasks with multiple labels?
-
January 2026
New paper at Computer Science Review: An overview of sign language processing from natural language processing perspective.
-
June 2025
New paper at Knowledge Based Systems: Modest: A dataset for multi domain scientific title generation.
-
June 2025
Joined Hacettepe University as an assistant professor of computer science.
Research
- Multimodal foundation models
- Integrating language, vision, motion and other behavioral signals into a single learning system instead of one modality at a time.
- Sign language processing
- Recognition, production, translation, and the linguistic structure that general-purpose video models tend to miss.
- Data-efficient & low-resource learning
- Representation learning that generalizes from little or no labelled data, for problems where annotation is scarce or impossible to collect at scale.
- Socially intelligent AI
- Systems that pick up on the behavioral signals underlying human interaction, not just its surface content.
Earlier work covered production machine learning systems for video surveillance.
Selected publications
-
2026
An overview of sign language processing from natural language processing perspective
Computer Science Review
An NLP-centred survey of sign language processing: which tasks exist, which are missing, what it shares with NLP, and where linguistically grounded work should go next.
-
2025
MoDeST: a dataset for multi-domain scientific title generation
Knowledge-Based Systems
A multi-domain, English–Turkish dataset for generating scientific titles, with an evaluation of LLMs across learning settings and evidence for domain-aware modelling.
-
2024
Cross-lingual few-shot sign language recognition
Pattern Recognition
Recognising signs across languages from a few examples, using textual and attribute-based class representations together with visual features.
-
2022
Towards zero-shot sign language recognition
IEEE Transactions on Pattern Analysis and Machine Intelligence
An embedding framework built on spatio-temporal video features and hand landmarks, with multilingual benchmarks for signs the model never saw in training.
-
2021
IEEE/CVF Winter Conference on Applications of Computer Vision
WildestFaces, plus a clean-to-violent domain transfer method for identifying people in footage that breaks standard face recognition.
-
2019
Zero-shot sign language recognition: can textual data uncover sign languages?
British Machine Vision Conference
The paper that introduced zero-shot sign language recognition, pairing written sign descriptions with a spatio-temporal video model.
Teaching
- AIE635 Reinforcement LearningInstitute of Informatics, Hacettepe University · Spring 2025–26
- AIE683 Machine LearningInstitute of Informatics, Hacettepe University · Spring 2025–26
- COM4513 Special Topics I: Introduction to Computer VisionComputer Engineering, Ankara University · Fall 2024–25
- COM3548 Pattern RecognitionComputer Engineering, Ankara University · Spring 2023–24
- COM4523 Artificial Neural NetworksComputer Engineering, Ankara University · Fall 2023–24
Patents
- Red light violation detection on road-facing in-car camerasY. C. Bilge et al. · 2023 003762 · registered 25.12.2025
- A method for real-time running detection and tracking in video footageY. C. Bilge et al. · 2023 013301 · registered 21.02.2025
- Calibration of angle measuring sensors (IMU) in portable devices using DEM and landform signatureY. C. Bilge et al. · 2023 003176 · registered 21.02.2024