Lucas Martini



Lucas Martini

I am a researcher working in the field of computational neuroscience, computer vision and machine learning, with a focus on 3D pose estimation, motion capture, animal behavior analysis, and representation learning. After I received my Bachelor's and Master's Degree at the Karlsruhe Institute of Technology (KIT), I started my PhD at the University of Tübingen and the International Max-Planck Research School for Intelligent Systems (IMPRS-IS).

Recently, I have been exploring how visual systems encode motion, shape, and social interaction. Please find a list of recent projects and publications that I have been involved in below.

Profile photo

Projects

BigMaQ teaser

BigMaQ: A Big Macaque Motion and Animation Dataset

Lucas Martini, Alexander Lappe, Anna Bognár, Rufin Vogels, Martin A. Giese
ICLR, 2026

A large-scale 3D motion dataset of rhesus macaques with detailed pose annotations, showcasing improved action recognition over classical behavioral descriptors in animals using surface based modeling.

Pose tuning paper teaser

Keypoint-based modeling reveals fine-grained body pose tuning in superior temporal sulcus neurons

Rajani Raman, Anna Bognár, Ghazaleh Ghamkhari Nejad, Albert Mukovskiy, Lucas Martini, Martin A. Giese & Rufin Vogels
Nature Communications, 2025

Investigates the body pose and orientation tuning of neurons based on avatar stimuli and keypoint modeling, deepening our understanding of body pose representations in the brain.

Shared-feature visualization teaser

Parallel backpropagation for shared-feature visualization

Alexander Lappe, Anna Bognár, Ghazaleh Ghamkhari Nejad, Albert Mukovskiy, Lucas Martini, Martin A. Giese, Rufin Vogels
NeurIPS, 2024

A deep-learning feature visualization approach to identify shared features of body selective neurons with associated out-of-category object features.

MacAction teaser

MacAction: Realistic 3D macaque body animation based on multi-camera markerless motion capture

Lucas Martini, Anna Bognár, Rufin Vogels, Martin A. Giese

Investigates the realism of full-body avatar animations based on keypoints and deep-learning based temporal interpolation.