YORU Tracker is installed with uv, next to a YORU checkout. One uv sync builds an environment that holds both: YORU (installed editable from the folder beside it) and YORU Tracker on top.
YORU Tracker is developed for Windows 10 or later, and its CI runs on Windows.
Detection is done by YORU, so it has YORU’s needs. To detect on a GPU, install an NVIDIA driver: YORU v2.0.0 Beta 4 uses the CUDA 12.8 build of PyTorch, which needs driver 570 or newer (572.xx on Windows) and covers every card up to the RTX 50-series. The CUDA toolkit is not needed: the PyTorch wheels carry their own CUDA runtime. Check the driver with:
nvidia-smi
Without a usable GPU, detection runs on the CPU — everything works, but more slowly. Tracking itself always runs on the CPU and is light. See YORU’s Beta Install guide for the details of drivers and compute devices.
Install uv and Git. In PowerShell:
winget install --id=astral-sh.uv -e
winget install --id=Git.Git -e
uv also installs Python 3.10 for you. You do not need conda or a separate Python.
YORU Tracker finds YORU at ../YORU, so both repositories must be in the same parent folder:
YORU-dev/
├── YORU/ # https://github.com/Kamikouchi-lab/YORU
└── YORU-Tracker/ # https://github.com/Kamikouchi-lab/YORU-Tracker
mkdir YORU-dev
cd YORU-dev
git clone -b v2.0.0-beta.4 https://github.com/Kamikouchi-lab/YORU.git
git clone -b v0.1.0 https://github.com/Kamikouchi-lab/YORU-Tracker.git
Which YORU? YORU Tracker requires YORU 2.0.0 Beta 4 or later (and below 3): Beta 4 is the first YORU release with the external API that YORU documents for YORU Tracker, and the version YORU Tracker’s CI tests against. The stable YORU v1.1.x is too old: with it beside YORU Tracker, uv sync fails.
If you already have a YORU folder, you can use it: check out the release there (git -C YORU fetch --tags, then git -C YORU checkout v2.0.0-beta.4) instead of cloning again.
cd YORU-Tracker
uv sync
This creates YORU-Tracker\.venv and installs YORU with its CUDA build of PyTorch, then YORU Tracker. The first sync downloads PyTorch and takes a while.
uv run yoru-tracker --version
prints yoru-tracker 0.1.0.
uv run yoru-tracker
uv run python -m yoru_tracker does the same. The start screen opens; continue with Video Tracking.
Do not run a bare python on Windows. It may be the Microsoft Store stub, not the environment’s Python. Use uv run ..., or the environment’s interpreter directly: .venv\Scripts\python.exe -m yoru_tracker.
Both folders are checked out at a release tag, so update them by checking out the newer tags, then sync again from the YORU Tracker folder. Each YORU Tracker release names the YORU version it needs:
cd YORU-dev
git -C YORU fetch --tags
git -C YORU checkout v2.0.0-beta.4
git -C YORU-Tracker fetch --tags
git -C YORU-Tracker checkout v0.1.0
cd YORU-Tracker
uv sync
Replace the two tags with the ones the release you are moving to names. git pull does not work on a tag checkout.
| File | Content |
|---|---|
~/.yoru/logs/yoru_tracker.log |
Every error shown in the window (with its traceback), failed batch files, tracker resets, configuration loads. Next to YORU’s own yoru.log. Rotated at 5 MB, with three old files kept. |
~/.yoru/yoru_tracker_gui.json |
What the window remembers between sessions: model, detector threshold, folders, camera ID, tracker settings. Delete it to start from the defaults. |
On Windows, ~ is %USERPROFILE%, so the log is %USERPROFILE%\.yoru\logs\yoru_tracker.log. Both files move with the YORU_HOME environment variable, as YORU’s do. Help → Open log folder opens the folder.
Tracking results are written where you choose; see Output Files.
The detector is loaded by YORU, so YORU decides where it runs: CUDA, then Apple MPS, then the CPU. To choose yourself, set YORU_DEVICE before launching, for example in PowerShell:
$env:YORU_DEVICE = "cpu"
uv run yoru-tracker
See Choosing the compute device in the YORU documentation.