Mohammad Asim

I am a PhD student at Max-Planck-Institut für Informatik working under the supervision of Jan Eric Lenssen. My interests lie at the intersection of 3D Computer Vision/Graphics and Diffusion-based Generative models for efficient 3D scene reconstruction and generation.

I recently finished my master's degree in Computer Science from Universität des Saarlandes in 2024 specializing in 3D computer vision and machine learning. For my thesis, I joined Geometric Representation Learning group at Max-Planck-Institut für Informatik advised by Dr. Jan Eric Lenssen and co-advised by Christopher Wewer. My thesis primarily focused on performing diffusion directly on the parameters of 3D Gaussian representation for object and scene level reconstruction from single or few images.

Email / Github / LinkedIn / Google Scholar

Research Activities

Updates

[25.02.2026] Our recent work SceneTok is accepted to CVPR 2026🎉!
[26.02.2025] Our recent work MEt3R is accepted to CVPR 2025🎉!
[10.01.2025] Project website and arxiv version for MEt3R is up!
[06.08.2024] Defended my masters' thesis ⚔️!
[01.08.2024] Joined as a PhD student at Geometric Representation Learning (Max-Planck-Institute for Informatics)

Publications

	SceneTok: A Compressed, Diffusable Token Space for 3D Scenes Mohammad Asim, Christopher Wewer, Jan Eric Lenssen IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026 project page / arXiv A novel tokenizer encoding 3D scenes into a compressed, unstructured and diffusable 1D token set to allow for efficient novel-view synthesis and generation.
	MEt3R: Measuring Multi-View Consistency in Generated Images Mohammad Asim, Christopher Wewer, Thomas Wimmer, Bernt Schiele, Jan Eric Lenssen IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025 project page / arXiv A differentiable metric to measure multi-view consistency between an image pair.

Copied from source code.