Computational tools for studying complex biological systems
The lab of Talmo Pereira at the Salk Institute for Biological Studies builds deep learning and computer vision methods to capture biological motion, phenotype animals and plants at scale, and emulate living systems in simulation.
- Led by
- Talmo Pereira
- Institute
- Salk Institute for Biological Studies
La Jolla, California - Founded
- 2021
- People
- 18 in the lab, 48 alumni
- Software
- Open source, on GitHub
Research
All research- a Biological motion capture Pose estimation, identity tracking and 3D: from video to how every animal in it moves.
- b Phenoinformatics Systems for phenotyping at scale, from home cages recorded around the clock to the roots of crop plants.
- c Virtual biology Computational models that emulate biological systems closely enough to run experiments in simulation.
Recent publications
Papers, preprints and proceedings. A preprint that was later published is listed once.
All 40 publications-
LUC3D: Multi-animal 3D pose tracking in the browser
bioRxiv · · preprint
-
Shaped by What's Missing: Topological Invariance Simplification Discovers Behavioral Transitions
NeurIPS 2026 Workshop: NeurReps ·
-
Discrete Actions for Naturalistic Behavior in Embodied Biomechanical Animal Models
NeurIPS 2026 Workshop: Neuro-Symbolic Embodied Intelligence ·
-
A latent control model for realistic rodent motion
NeurIPS 2026 ·
People
Scientists, engineers and students from neuroscience, computer science, biology and beyond: 18 people in the lab today, and 48 alumni.
Meet the team
Software
Open-source tools we build and maintain, starting with SLEAP for multi-animal pose tracking.
All tools- SLEAP A deep learning framework for multi-animal pose tracking.
- sleap-nn The neural network backend for training and inference, split out of SLEAP so the models can be developed and used independently of the labelling GUI.
- sleap-io Standalone utilities for reading, writing and converting pose tracking data.
- sleap-app A web-based labelling GUI for SLEAP, running in the browser rather than as a desktop install.
- DREEM DREEM Relates Every Entity's Motion — global tracking that links detections into identities across a whole video rather than frame to frame.
Join us
We place a high value on mentorship, work-life balance and open communication, and we recruit at every level: engineers, postdocs, graduate students and interns.