☕

Marc Benedí San Millán

PhD Candidate @ Visual Computing Group

Technische Universität München

👋 Hola! Hello! Hallo!

I am a PhD candidate at Matthias Nießner’s Visual Computing Group, at the Technical University of Munich.

I’m interested in computer vision, deep learning and optimization. My current research interests are focused on virtual avatars, generative models and 3D scene representations.

Before joining the group, I received a Master’s Degree in Computer Science from the Technical University of Munich and a Bachelor’s Degree in Computer Science from the Polytechnic University of Catalonia.

Publications

Quickly discover relevant content by filtering publications.

Animating the Uncaptured: Humanoid Mesh Animation with Video Diffusion Models

We present a novel approach to generating realistic 4D animations from static 3D humanoid meshes using generative video models. By leveraging powerful motion priors, our method transforms a static 3D mesh into an animated sequence based on a text prompt. We optimize the animation using an SMPL-based motion representation, enabling a cost-effective and accessible solution for creating diverse and lifelike character animations.

Marc Benedí San Millán, Angela Dai, Matthias Nießner

Animating the Uncaptured: Humanoid Mesh Animation with Video Diffusion Models

Experience

PhD Candidate

Visual Computing & Artificial Intelligence Lab, Technische Universität München

January 2023 – Present Munich, DE

Research Fellowship

Visual Computing & Artificial Intelligence Lab, Technische Universität München

April 2021 – March 2022 Munich, DE

Fellowship awarded to outstanding Master students to allow them developing state-of-the-art models. See page.
See project’s page.

Working Student

Visual Computing & Artificial Intelligence Lab, Technische Universität München

April 2020 – September 2022 Munich, DE

Built a high-performance GPU cluster for the Deep Learning laboratory.
Built a frontend with Vue.js and a backend for the cluster’s users.

Software Engineer

HELM Mobile Development

September 2018 – September 2019 Barcelona, ES

Android development with Kotlin.
Backend development with Kotlin.
Frontend development with Vue.js

Summer Intern

European Organization for Nuclear Research (CERN)

June 2018 – September 2018 Geneva, CH

Designed and built a cluster for the BE-OP-LHC team at CERN using Kubernetes.

Research Assistant

Polytechnic University of Catalonia, Barcelona Tech

September 2017 – January 2018 Barcelona, ES

Worked on a novel conversion of Binary Decision Trees to Conjunctive Normal Forms.
See project’s page.

Summer Intern

Ernst & Young

June 2017 – September 2017 Barcelona, ES

Frontend and backend development.

Projects

Quickly discover relevant content by filtering projects or by clicking here.

Open Source Contributions

Summary of my Open Source contributions

Learning Robust Correspondences Estimation

An end-to-end differentiable model for relative camera pose estimation from RGB-D frame pairs. The network learns self-supervised confidence scores for predicted correspondences, which are then aligned by a differentiable weighted Procrustes solver. Evaluated on ScanNet with strong results in wide-baseline and occlusion-heavy regimes.

End-to-end Sparse Correspondence Prediction for Relative Camera Pose Estimation

We present an end-to-end learnable, differentiable method for pairwise relative pose registration of RGB-D frames. Our method is robust to big camera motions thanks to a self-supervised weighting of the predicted correspondences between the frames. Given a pair of frames, our method estimates matches of points and their visibility score. A self-supervised model predicts a confidence weight for visible matches. Finally, visible matches and their weight are fed into a differentiable weighted Procrustes aligner which estimates the rigid transformation between the input frames.

Kinect Fusion: Dense Surface Mapping and Tracking

Implementation of the paper “Kinect Fusion: Real-Time Dense Surface Mapping and Tracking”

SLAM for Autonomous Vehicles

Worked on the SLAM pipeline of an autonomous driving vehicle using ROS, C++, PCL, Ceres Solver, and Google Cartographer. Developed a ROS package that turns SLAM point clouds into a cost map encoding both occupancy and surface preference (e.g., asphalt over grass) for path planning.

Divergence-Free Shape Correspondence with Time Dependent Vector Fields

Extended Eisenberger et al.’s divergence-free shape interpolation by introducing time-dependent vector fields, lifting the limitation that different parts of a shape cannot move in conflicting directions through the same point in space. Correspondence and matching are jointly solved over the full sequence during optimization.

SLURM Cluster - Dashboard and other tools

Designed and deployed a SLURM cluster for the TUM Visual Computing Group, including LDAP-backed authentication, NFS storage, monitoring, and Ansible-driven node provisioning. Built a Vue.js + Flask dashboard so researchers can visualize cluster state, jobs, and resource usage, and released companion open-source tools (LDAP Triggers, LDAP Emails) used in day-to-day administration.

Design of an Environment for Solving pseudo-Boolean Optimization Problems

Built a C++ library and an efficient CNF encoding for solving Pseudo-Boolean Optimization problems via off-the-shelf SAT solvers, with linear and binary minimization strategies and configurable timeout policies for time-bounded solving.

StirHack 2016

Writing about StirHack16, the first hackathon I ever attended, brings me very good memories. In this page I will focus on the project itself, although some day I should write about the hackathon experience. 😃