DeepLabCut-live-GUI#
A graphical application for real-time pose estimation with DeepLabCut using one or more cameras.
This GUI is designed for scientists and experimenters who want to preview, run inference, and record synchronized video with pose overlays—without writing code.
Table of Contents#
Caution
Please be aware of the Current limitations
Description#
What this software does#
Live camera preview from one or multiple cameras
Real-time pose inference using DeepLabCut Live models
Multi-camera support with tiled display
Video recording (raw or with pose and bounding-box overlays)
Session-based data organization with reproducible naming
Optional processor plugins to extend behavior (e.g. remote control, triggers)
The application is built with PySide6 (Qt) and is intended for interactive experimental use rather than offline batch processing.
Typical workflow#
Install the application and required camera backends
Configure cameras (single or multi-camera)
Select a DeepLabCut Live model
Start preview and verify frame rate
Run pose inference on a selected camera
Record video (optionally with overlays)
With organized results by session and run
Each of these steps is covered in the Quickstart and User Guide sections of this documentation.
Who this is for#
Neuroscience and behavior labs
Experimentalists running real-time tracking
Anyone who wants a GUI-first workflow for DeepLabCut Live
Current limitations#
Before getting started, be aware of the following constraints:
Pose inference runs on one selected camera at a time (even in multi-camera mode)
Camera synchronization depends on backend capabilities and hardware
OpenCV controls for resolution and FPS are “best effort” and may not work with all cameras. Expect inconsistencies when setting certain frame rates or resolutions as resolution depends on the device driver.
DeepLabCut Live models must be exported and compatible with the selected backend
Some SuperAnimal models from The DeepLabCut Model Zoo! may not work out of the box.
This is currently a known issue for:SuperHuman model (missing detector)
Performance depends on camera resolution, frame rate, GPU availability, and codec choice
Expect bottlenecks with heavy models, multiple high-resolution cameras, or CPU-only inference.
Feedback, issues, and contributions#
This project is under active development. Feedback from real experimental use is highly valued.
Please report issues, suggest features, or contribute to the codebase on GitHub !