Christopher Wewer

Research Interests #

My interest lies in 3D computer vision, generative modeling, and everything in between. Currently, my research is focused on designing scalable generative models, e.g., diffusion and flow transformers for 3D scene reconstruction from sparse views and reasoning in spatial domains. Throughout various projects, I further gained expertise in 3D representations such as neural radiance fields and 3D Gaussian Splatting.

News #

• 
  SemanticNVS: Improving Semantic Scene Understanding in Generative Novel View Synthesis

  ICML 2026

  Xinya Chen, Christopher Wewer, Jiahao Xie, Xinting Hu, Jan Eric Lenssen

  Short Abstract: Extracting semantic features from given conditioning before generation improves the quality of generative novel view synthesis.

  Project Page Paper arXiv Code
• 
  SceneTok: A Compressed, Diffusable Token Space for 3D Scenes

  CVPR 2026

  Mohammad Asim, Christopher Wewer^†, Jan Eric Lenssen

  Short Abstract: A scene tokenizer that encodes and decodes view-sets of a scene to and from a highly compressed, 1D token set representation that can be used for efficient latent scene diffusion.

  Project Page Paper arXiv Code
• 
  Spatial Reasoners for Continuous Variables in Any Domain

  CodeML Workshop @ ICML 2025

  Bart Pogodzinski, Christopher Wewer, Bernt Schiele, Jan Eric Lenssen

  Short Abstract: A software framework to bring spatial reasoning networks (SRMs) to other domains. Allows reasoning over sets of continuous variables

  Project Page Paper arXiv Code
• 
  Spatial Reasoning with Denoising Models

  ICML 2025

  Christopher Wewer, Bart Pogodzinski, Bernt Schiele, Jan Eric Lenssen

  Short Abstract: A framework to propagate belief over a set of continuous variables (e.g. image patches) with generative denoising models. Explores the amount of sequentialization and generation order.

  Project Page Paper arXiv Code
• 
  MEt3R: Measuring Multi-View Consistency in Generated Images

  CVPR 2025

  Mohammad Asim, Christopher Wewer, Thomas Wimmer, Bernt Schiele, Jan Eric Lenssen

  Short Abstract: A metric that measures 3D consistency between two generated images without requiring camera poses. It uses DUSt3R to perform reconstruction and compares projected DINO features.

  Project Page Paper arXiv Code
• 
  Spurfies: Sparse Surface Reconstruction using Local Geometry Priors

  3DV 2025 (Oral Presentation, Award Candidate)

  Kevin Raj, Christopher Wewer, Raza Yunus, Eddy Ilg, Jan Eric Lenssen

  Short Abstract: A method leveraging synthetic data to learn local surface priors for surface reconstruction from few images. Can be applied to both bounded and unbounded scenes.

  Project Page Paper arXiv Code
• 
  latentSplat: Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction

  ECCV 2024

  Christopher Wewer, Kevin Raj, Eddy Ilg, Bernt Schiele, Jan Eric Lenssen

  Short Abstract: A fast autoencoder that encodes image pairs into 3D variational feature Gaussians that model uncertainty individually for different locations in 3D space. Trained purely on videos.

  Project Page Paper arXiv Code
• 
  Neural Parametric Gaussians for Monocular Non-Rigid Object Reconstruction

  CVPR 2024

  Devikalyan Das, Christopher Wewer, Raza Yunus, Eddy Ilg, Jan Eric Lenssen

  Short Abstract: A two-stage approach for non-rigid reconstruction from monocular videos. A coarse point template is found first to act as regularization for a local 3D Gaussians representation.

  Project Page Paper arXiv Code
• 
  Neural Point Cloud Diffusion for Disentangled 3D Shape and Appearance Generation

  CVPR 2024

  Philipp Schröppel, Christopher Wewer, Jan Eric Lenssen, Eddy Ilg, Thomas Brox

  Short Abstract: A denoising diffusion method on point clouds with features defining a radiance field. The disentangled representation allows for individual generation of shape and appearance.

  Project Page Paper arXiv Code
• 
  SimNP: Learning Self-Similarity Priors between Neural Points

  ICCV 2023

  Christopher Wewer, Eddy Ilg, Bernt Schiele, Jan Eric Lenssen

  Short Abstract: A renderable neural point radiance field that learns category-level self-similarities from data by connecting coherent neural points to embeddings via optimized bipartite attention scores.

  Project Page Paper arXiv Code

Academic Service #