Research Experience
- 02.2025 - present: Master’s Thesis Student in Computer Graphics Laboratory, ETH Zurich and Disney Research Studios
- Research on generative video compression using video diffusion models in conjunction with implicit neural representations to learn low-rate realistic video representations.
- Finetuning existing video diffusion models such as Wan2.1 and SVD to adapt their backbone for external video representation conditioning such as NeRV and HiNeRV for high fidelity and perceptually-pleasing videos.
- Supervisors: Lucas Relic and Roberto Azevedo
- 08.2024 - 01.2025: Research Assistant in Computer Vision and Geometry Group, ETH Zürich
- Research on generalizable sparse view 3D Gaussian splatting for 3D reconstruction.
- I worked on generalizable 3D Gaussian splatting for scene reconstruction. As novel view synthesis objective through regular photogrammetric loss is not suitable for surface reconstruction, I worked on adapting existing sparse view generalizable 3D Gaussian splatting methods to surface reconstruction via Gaussian Opacity Fields.
- Supervisors: Haofei Xu and Fangjinhua Wang
- 03.2024 - 06.2024: Graduate Research Student in Computer Vision Lab, ETH Zürich
- Research on learned 3D Gaussian splatting representation compression.
- Worked on 3D Gaussian splatting compression through the literature on learned image compression. Representing a scene with 3D Gaussian splatting requires a large number of Gaussian primitives, from hundreds of thousands to several millions, resulting in high storage complexity. My projects focused on direct adoption of learned image compression networks for Gaussian splatting compression through a short finetuning stage.
- Supervisors: Dr. Ertunç Erdil and Yannick Strümpler
- 06.2020 - 09.2022: Undergraduate Research Assistant in KUIS AI Lab, Koç University
- Research on learned image/video compression. Work accepted in ICIP 2022.
- Worked on image and video compression using recurrent and fully convolutional variational auto-encoders. My projects included our work on the design of a novel bidirectional frame compression network and several implementation of recent works in the literature using PyTorch machine learning framework.
- Supervisor: Professor A. Murat Tekalp
- Research interests of Professor Tekalp captures a wide range of image/video processing applications including image/video super resolution, compression and restoration. With his vast research experience, Professor Tekalp helped me broaden my knowledge on above mentioned tasks and guided me through our research.
- 06.2021 - 09.2021: Summer Research Assistant in KUIS AI Lab, Koç University
- Studied style transfer and image manipulation with generative adversarial networks.
- Worked on implementation of “Swapping Autoencoder for Deep Image Manipulation” by Park et al. (2020) and presented the work together with a demo.
- Supervisor: Professor Aykut Erdem
- Professor Erdem conducts research on a diverse set of topics, ranging from image editing to visual saliency estimation, and to multimodal learning for integrated vision and language.
