Christian Schmidt

Hi, I'm Christian! I’m a PhD student in the Computer Vision Group at RWTH Aachen University, supervised by Prof. Bastian Leibe.

My research focuses on 3D reconstruction from images: Gaussian splatting, camera pose estimation, and making feed-forward multi-view geometry models fast and scalable.

Publications

An indoor scene reconstructed with colored 3D motion tracks.

arXiv 2026

ARROW: Arbitrary Reconstruction and Tracking of 4D Observations in the Wild

Ilya Fradlin, Christian Schmidt, Jens Piekenbrinck, Karim Knaebel, Gonzalo Martin Garcia, Bastian Leibe

Joint 4D reconstruction and point tracking from arbitrary image sets, across viewpoints and time.

Paper Code Project

A SurGe point map of a vintage car, shown half in RGB and half as surface normals.

NeurIPS 2026

SurGe: Improved Surface Geometry in Point Maps

Karim Knaebel, Gonzalo Martin Garcia, Christian Schmidt, Ilya Fradlin, Lucas Nunes, Daan de Geus, Bastian Leibe

A monocular point map model with a point gradient matching loss and a Neighborhood Attention Decoder for sharper local surface geometry.

Paper Code Project

Plot showing global attention dominating VGGT runtime as the number of frames grows.

CVPR 2026

Block-Sparse Global Attention for Efficient Multi-View Geometry Transformers

Chung-Shien Brian Wang*, Christian Schmidt*, Jens Piekenbrinck, Bastian Leibe

An analysis of feed-forward multi-view transformers like VGGT, π³, and MapAnything, resulting in a training-free, block-sparse global attention variant with up to 3× faster inference on large image sets.

Paper Code Project

A tabletop scene split into its RGB rendering and the PCA projection of the learned per-Gaussian features.

CVPR Workshops 2025

OpenSplat3D: Open-Vocabulary 3D Instance Segmentation using Gaussian Splatting

Jens Piekenbrinck, Christian Schmidt, Alexander Hermans, Narunas Vaskevicius, Timm Linder, Bastian Leibe

Open-vocabulary 3D instance segmentation in Gaussian splatting via feature splatting, SAM-guided contrastive learning, and vision-language embeddings.

Paper Code Project

Example images with their predicted depth and surface normal maps.

WACV 2025 · Oral

Fine-Tuning Image-Conditional Diffusion Models is Easier than You Think

Gonzalo Martin Garcia, Karim Abou Zeid, Christian Schmidt, Daan de Geus, Alexander Hermans, Bastian Leibe

A fix to the inference pipeline makes diffusion-based depth estimators 200× faster; end-to-end fine-tuning then yields strong single-step depth and normal estimation.

Paper Code Project

Novel view rendering of a museum scene reconstructed without accurate camera poses.

IROS 2024 · Oral

Look Gauss, No Pose: Novel View Synthesis using Gaussian Splatting without Accurate Pose Initialization

Christian Schmidt, Jens Piekenbrinck, Bastian Leibe

Jointly optimizing camera extrinsics and 3D Gaussians from photometric residuals, with analytical pose gradients integrated into the CUDA rasterizer.

Paper Code

Sampling patterns of a regular 3D convolution, a deformable 3D convolution, and a dynamic dilated 3D convolution.

WACV 2022

D²Conv3D: Dynamic Dilated Convolutions for Object Segmentation in Videos

Christian Schmidt, Ali Athar, Sabarinath Mahadevan, Bastian Leibe

Spatio-temporal dynamic dilated convolutions as a drop-in replacement for 3D convolutions, achieving state of the art (at the time) on DAVIS 2016 unsupervised VOS.

Paper Code

Contact

schmidt@vision.rwth-aachen.de

Computer Vision Group, RWTH Aachen University