https://github.com/driscollh/opencap-offline
https://github.com/driscollh/opencap-offline
biomechanics biomechanics-analysis motion-capture opencap opensim
Last synced: 3 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/driscollh/opencap-offline
- Owner: driscollh
- License: apache-2.0
- Created: 2026-03-13T05:18:59.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-04-06T06:55:43.000Z (3 months ago)
- Last Synced: 2026-04-06T08:45:38.671Z (3 months ago)
- Topics: biomechanics, biomechanics-analysis, motion-capture, opencap, opensim
- Language: Python
- Homepage:
- Size: 19.7 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Funding: .github/FUNDING.yml
- License: LICENSE
Awesome Lists containing this project
README
# OpenCap-Offline
A fully offline, localized implementation of Stanford NMBL's OpenCap. This version adapts the original cloud-based processing pipeline to run entirely on a local machine, ensuring strict data privacy and allowing for use in environments without internet access.
This version of the software requires a sufficiently powerful GPU, but cannot run on blackwell architecture cards (i.e. Nvidia 50-series cards).
I recommend use with Nvidia 30-series or 40-series cards with minimum 12GB graphics ram.
[](https://www.buymeacoffee.com/driscollh)
## Prerequisites
* Python 3.9.25
* Git
* NVIDIA GPU (RTX 3000 series or 4000 series recommended) - this setup is not yet suitable for GPUs with Blackwell architecture
* Anaconda or Miniconda
## Installation & Setup
**1. Download the Code**
Clone this repository to your local machine:
`git clone https://github.com/driscollh/opencap-offline.git`
**2. Create a Python Environment with OpenSim**
conda create -n opencap_env python=3.9 -y
conda activate opencap_env
conda install -c opensim-org opensim=4.4 -y
**3. Install Python Packages**
Navigate into the main folder and install the required Python environment packages:
`pip install -r requirements.txt`
**4. Download the Local Dependencies (Required)**
Because the machine learning models and background engines (OpenPose, FFmpeg) are too large for GitHub, they are hosted securely on Zenodo.
* Go to the V3 Zenodo archive: [https://doi.org/10.5281/zenodo.19447679](https://doi.org/10.5281/zenodo.19447679)
* Download the `dependencies.zip` file.
* Extract the contents directly into the empty `dependencies/` folder inside this project. This should result in three separate subfolders, one for each key dependency.
* Extract the `RTMPose.zip` file directly into the main project folder next to `pyqt5_launcher_improved.py`. This folder should be titled "Blackwell_RTMPose", and contain a `dependencies/` folder and `mmcv/` folder.
## Usage
Once the environment is set up and the dependencies are in place, you can launch the local processing by running the main Python script in your environment:
`python pyqt5_launcher_improved.py`
or
`python simple_launcher.py`
## VideoRecording
Unlike the online version of OpenCap, this offline version requires additional video capture for calibration of camera intrinsics and extrinsics.
Please consult the `recording_practices.pdf` file prior to data collection.
## Acknowledgments and Licensing
This project is built upon the original open-source OpenCap project by the Stanford Neuromuscular Biomechanics Laboratory.
* **Original OpenCap:** [https://github.com/opencap-org/opencap-core](https://github.com/opencap-org/opencap-core)
* **License:** Distributed under the Apache 2.0 License.