DeepLabCut
Markerless pose estimation for any animal including humans: label a few hundred frames, train your own tracker.
About
DeepLabCut brought pose estimation to biology labs that had never touched deep learning. You label a few hundred frames in its GUI, and the active-learning loop trains a network that tracks user-defined body parts in new videos with accuracy rivaling human labelers. Multi-animal projects (v2.2) add identity and tracking; the v3.0 rewrite moved the backend to PyTorch.
The Model Zoo and the newer SuperAnimal foundation models offer zero-shot pose across species, though those pretrained weights are research-use only. The core package stays LGPL-3.0, and it remains the standard tool for markerless tracking in neuroscience and behavioral research.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-claimed (Nature Neuroscience 2018): test error comparable to human labeling accuracy with ~200 training frames; SuperAnimal (Nature Communications 2024) reports strong cross-species zero-shot transfer. Both provider-claimed; no independent product benchmarks were found.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 9/10
Scored 9/10 (verified, metrics-v1). Evidence: peer-reviewed Nature Neuroscience (2018), Nature Protocols (2019), and Nature Communications (2024, SuperAnimal) papers by the named lab; LGPL-3.0 license; very large research adoption and active maintenance (v3.0 PyTorch rewrite). Strengths: genuine openness, peer review, sustained development, huge citation base. Docked for: (1) SuperAnimal pretrained weights are non-commercial, a carve-out on openness; (2) benchmark figures are provider-claimed only. License: LGPL-3.0 (core); SuperAnimal models research-use only.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 10.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 9.0/10
- · Mathis Lab; Nature Neuroscience 2018, Nature Protocols 2019, Nature Communications 2024 (SuperAnimal).
- · Provider-claimed ~human-level labeling accuracy with ~200 training frames; v3.0 = PyTorch rewrite.
- · Standard tool for markerless animal pose; SuperAnimal non-commercial carve-out noted.