I am a PhD Candidate at the University of Goettingen, working on self-supervised video understanding models for real world applications.
I build video understanding models that work on real-world footage with limited labeled data. What drives me is helping scientists reach reliable insights faster. That is why my main application is animal behavior recognition, where I help ecologists to accelerate and scale their data analysis. To build models that generalize across varying settings, I combine self-supervised video representation learning, large-scale pretraining, and domain adaptation. Previously I worked on modeling human motion and interactions, and on LLM evaluation.
🌐 felixbmuller.github.io · Working at Eckerlab, University of Goettingen
| Year | Work | Venue | Repository |
|---|---|---|---|
| 2026 | PriVi — Towards a General-Purpose Video Model for Primate Behavior in the Wild | CVPR | privi |
| 2026 | TrAction — Action Recognition with Sparse Trajectories | arXiv | TrAction |
| 2025 | Domain-Adaptive Pretraining Improves Primate Behavior Recognition | CV4Animals @ CVPR (Oral) | dap-behavior |
| 2024 | Massively Multi-Person 3D Human Motion Forecasting with Scene Context | ABAW @ ECCV | SAST |
| 2024 | LLMs and Memorization — On Quality and Specificity of Copyright Compliance | AIES | llms-memorization-copyright |
| 2023 | Humans in Kitchens — A Dataset for Multi-Person Motion Forecasting | NeurIPS D&B | hik |
Full list on my website and Google Scholar.
Website · Google Scholar · LinkedIn · Email