Installation

The current stable release is v0.4.43 (opens in a new tab). Python installations require Python 3.11 or newer; Python 3.12 is recommended.

Create a dedicated AnyLabeling environment. ONNX Runtime CPU and hardware-specific distributions replace one another and should not be mixed in the same environment.

1. Download a binary

Use the Download page for direct links, or choose the correct artifact from the latest GitHub release (opens in a new tab):

The six v0.4.43 artifacts publish SHA-256 digests and were launch-smoke-tested on native Linux, Windows, and macOS runners before upload. Do not use the v0.4.40 macOS or Linux artifacts; they were superseded because of packaging defects.

On Linux, make the downloaded file executable:

chmod +x AnyLabeling-Linux-CPU-x64
./AnyLabeling-Linux-CPU-x64

On macOS, extract the ZIP without flattening or replacing its symlinks, then launch the included binary:

ditto -x -k AnyLabeling-macOS-CPU.zip .
./AnyLabeling-macOS-CPU/anylabeling

Use the GPU archive instead to enable CoreML. If macOS blocks the first launch, open System Settings → Privacy & Security and allow the application.

2. Install from PyPI

Create and activate an isolated environment:

conda create -n anylabeling python=3.12
conda activate anylabeling

CPU

python -m pip install anylabeling
anylabeling

NVIDIA CUDA on Linux or Windows

Use the GPU distribution in a fresh environment:

python -m pip install anylabeling-gpu
anylabeling

The GPU distribution includes pip-managed CUDA 12 and cuDNN runtime libraries. A compatible NVIDIA driver is required, but a system CUDA toolkit is not.

Apple Silicon CoreML

Install Qt through Conda and the macOS extra through pip:

conda install -c conda-forge pyqt=6
python -m pip install "anylabeling[macos]"
export ANYLABELING_DEVICE=COREML
anylabeling

See Hardware Acceleration for DirectML, OpenVINO, TensorRT, and NPU setup.

3. Install from source

git clone https://github.com/vietanhdev/anylabeling.git
cd anylabeling
python -m pip install -e .
anylabeling

For NVIDIA CUDA development, replace the final install command with:

python -m pip install -e ".[gpu]"