Lightning pose app 2.0

Status: In development

• Multiview prediction viewer is in a usable state, needs polish.
• Labeler is in a usable state, needs polish.
• Model management is in development...

Features

The app has three modules:

1. Viewer: support for scrubbing through synced multiview video and their error traces
2. Labeling: support for multiview with camera calibrations, no longer using LabelStudio
3. Model management

Installation

You can install the app either locally or to a remote server.

Using environment managers like conda

We recommend having a single lightning-pose environment (say, lp) for all your lightning-pose related package installations. This keeps lightning-pose dependencies separate from any other data analysis stacks you may be using, while allowing the various lightning-pose packages to share dependencies.

A conda example:

# Create the environment
conda create -n lp python=3.12

# Activate the environment
conda activate lp

Installation

pip install lightning-pose lightning-pose-app

Installation for developers

# Install lightning-pose core
git clone https://github.com/paninski-lab/lightning-pose.git
cd lightning-pose
pip install -e ".[dev]"
cd ..

# Install the app
git clone https://github.com/paninski-lab/lightning-pose-app.git
cd lightning-pose-app/app_server
pip install -e .

Usage

1. First create a config file at ~/.lightning-pose/project.toml

data_dir = "/Path to a directory containing all the data (videos, labels, etc)."
model_dir = "/Path to a directory containing all the models."
views = [
    "topLeft",
    "(Your video filenames must contain a view name, ie session123_topLeft.mp4)",
]
keypoint_names = [
    "leftPaw",
    "rightPaw",
]

(This step will be automated in the future, but for now it's manual.) 2. Run the app: litpose app (equivalent to running uvicorn litpose_app.main:app --port 8080 if running without lightning-pose) 3. The webserver is now listening at http://localhost:8080!