Skip to content

SLEAP-NN

SLEAP pose estimation demo

CI codecov GitHub stars Release PyPI - Python Version PyPI

SLEAP-NN is the PyTorch backend for SLEAP, providing neural network training and inference for multi-animal pose estimation. It offers an end-to-end workflow from labeled data to tracked predictions, with seamless integration into SLEAP's GUI and command-line tools.

✨ Features

  • Multiple model types – Single instance, top-down, bottom-up, and centroid models
  • Modern backbones – UNet, ConvNeXt, and Swin Transformer architectures
  • Multi-GPU training – PyTorch Lightning with DDP support
  • Production export – ONNX and TensorRT for fast deployment
  • Flexible config – Hydra/OmegaConf for reproducible experiments

Let's start SLEAP-NNing! 🐭🐭

📚 Explore the Docs

  • 🚀 Quick Start

    Install and train your first model in 5 minutes.

    Get Started

  • 📖 Tutorials

    Quick start, first model tutorial, and example notebooks.

    Learn

  • 📚 Guides

    Training, inference, and tracking workflows.

    Explore

  • ⚙️ Configuration

    Customize your training config.

    Configure

  • 💻 CLI Reference

    Command-line interface documentation.

    Commands

  • 🎛️ Config Picker

    Generate training configs interactively in your browser.

    Try it now

🔄 Coming from SLEAP < v1.5?

Load legacy models for inference

Convert legacy config

SLEAP < v1.5 SLEAP-NN
TensorFlow/Keras PyTorch/Lightning
Single GPU Multi-GPU (DDP)
Limited export ONNX + TensorRT

Get Help