Who am I?

Hello! I’m Eren Çetin, a PhD student in the Photogrammetry and Remote Sensing (PRS) group at ETH Zürich, working on efficient video generation. My research focuses on video diffusion models, autoregressive video generation, and stateful generation, with the goal of making long, high-quality video synthesis fast and scalable.

Before starting my PhD, I completed my Master’s at ETH Zürich, specializing in signal processing, machine learning, and computer vision. I did my Master’s thesis at Disney Research Studios on generative video compression using video diffusion models and implicit neural representations (INRs).

My research broadly lies at the intersection of 3D vision, neural rendering, and generative models, aiming to build scalable and efficient 3D and video representation systems. I previously worked on generalizable 3D reconstruction from sparse views and compression of 3D Gaussian splatting representations.

I completed my Bachelor’s degrees in Electrical & Electronics Engineering and Economics at Koç University, graduating as salutatorian. My professional interests include video generation, video diffusion, 3D vision, 3D reconstruction, neural rendering, video compression, and deep generative modeling.

I primarily work with Python, PyTorch, CUDA, C/C++, and TypeScript, and enjoy bridging theory with large-scale implementations and visualization.

✨ Featured

SMART: Long-horizon Video Generation with Slot Memory Addressed By Region Tokens

SMART is a slot memory addressed by region tokens for long-horizon video generation. It keeps memory bounded so generation can continue indefinitely, while still recalling what was seen at a place when the camera returns.

View project page