Media1
Media2

AnyCam: Learning to Recover Camera Poses and Intrinsics from Casual Videos

CVPR, 2025

AnyCam is a fast transformer model that directly estimates camera poses and intrinsics from a dynamic video sequence in feed-forward fashion. This network can learn strong priors over realistic camera motion, by training on diverse, unlabelled video datasets obtained mostly from YouTube.

Media2

Nonisotropic Gaussian Diffusion for Realistic 3D Human Motion Prediction

CVPR, 2025

SkeletonDiffusion is a novel nonisotropic diffusion approach for 3D Human Motion Prediction, and the first computer vision method to show to use nonisotropic diffusion. We generate diverse and realistic motions achieving state-of-the-art performance.

Media1
Media2

Ground-Aware Automotive Radar Odometry

ICRA, 2025

We propose a simple, yet effective, heuristic-based method to extract the ground plane from single radar scans and perform ground plane matching between consecutive scan.

Media1
Media2

Lightspeed Computation of Geometry-aware Semantic Embeddings

TBD, 2025

A novel, optimal-transport based learning method to solve the challenge of matching semantically similar parts distinguished by their geometric properties, e.g., left/right eyes or front/back legs. It is faster and outperforms previous supervised methods in terms of semantic matching and geometric understanding.

Media1
Media2

Gaussian Splatting in Style

GCPR, 2024

We are the first to employ Gaussian Splatting to solve the task of scene stylization, extending the work of neural style transfer to three spatial dimensions.

Media1
Media2

Boosting Self-Supervision for Single View Scene Completion via Knowledge Distillation

CVPR, 2024

We use multi-view scene completion to supervise single-view scene completion and boost its performance. We propose both a novel multi-view scene completion network and a corresponding knowledge distillation scheme.

Media1
Media2

S4C: Self-Supervised Semantic Scene Completion with Neural Fields

3DV Spotlight, 2024

S4C is the first self-supervised approach to the Sematic Scence Completion task. It achives close to state-of-the-art performance on the KITTI-360 SSCBench dataset.

Media1

Learning Correspondence Uncertainty via Differentiable Nonlinear Least Squares

CVPR, 2023

A differentiable nonlinear least squares framework to account for uncertainty in relative pose estimation from feature correspondences regardless of the feature extraction algorithm of choice.

Media1
Media2

Probabilistic Normal Epipolar Constraint for Frame-To-Frame Rotation Optimization under Uncertain Feature Positions

CVPR, 2022

We propose a probabilistic extension to the normal epipolar constraint (NEC) which we call the PNEC. It allows to account for keypoint position uncertainty in images to produce more accurate frame to frame pose estimates.