David Recasens

PhD researcher in 3D Computer Vision and Machine Learning with background in Gaussian Splatting, 3D reconstruction, depth estimation, SLAM and generative AI. Experienced across top research labs (Disney, Meta, Huawei, ETH CVG) driving novel methods in AR/VR, robotics and medical imaging.

PhD under the guidance of Prof. Javier Civera at the Robotics Lab in the Universidad de Zaragoza. Research stay at Computer Vision and Geometry Group (CVG) directed by Prof. Marc Pollefeys in ETH Zurich, Switzerland, supervised by Martin. R. Oswald from September 2021 to March 2022. 3 Research Scientist internships in Zurich, Switzerland, at Huawei Research Center (Jan–Jul 2024), Meta Reality Labs (Jul 2024–Mar 2025) and Disney Research (Mar–Jun 2025).

Email  /  LinkedIn  /  Google Scholar  / 

profile photo
3D Reconstruction with 3D Gaussian Splatting
3DV 2026

Paper under review.

The Drunkard’s Odometry: Estimating Camera Motion in Deforming Scenes
David Recasens, Martin R. Oswald, Marc Pollefeys, Javier Civera
NeurIPS, 2023
Project page / arXiv paper / NeurIPS paper / Poster / Video / Code

The Drunkard’s Dataset, a challenging collection of synthetic data targeting visual navigation and reconstruction in deformable environments. And the Drunkard’s Odometry, a novel monocular RGB-D deformable odometry method that breaks down optical flow estimate into rigid-body camera motion and non-rigid scene deformation.

On the Uncertain Single-View Depths in Endoscopies
Javier Rodríguez-Puigvert, David Recasens, Javier Civera, Rubén Martínez-Cantín
MICCAI, 2022
Project page / MICCAI2022 paper / arXiv paper / Video demo

Deepening for the first time in Bayesian deep networks for single-view depth estimation in colonoscopies.

Endo-Depth-and-Motion: Reconstruction and Tracking in Endoscopic Videos using Depth Networks and Photometric Constraints
David Recasens, José Lamarca, José M. Fácil, José María M. Montiel,
Javier Civera
RA-L and IROS, 2021
Project page / RA-L paper / arXiv paper / Video demo / Code

A pipeline that estimates the 6-degrees-of-freedom camera pose and dense 3D scene models from monocular endoscopic videos.

Endomapper dataset of complete calibrated endoscopy procedures
Pablo Azagra, Carlos Sostres, Ángel Ferrández, David Recasens, et al.
Scientific Data (Nature), 2023
Project page / paper

First collection of complete endoscopy sequences acquired during regular medical practice to support the development visual SLAM methods.



Based on the Jon Barron's template.