pixano 0.7.2

Data-centric AI building blocks for computer vision applications

Data-centric AI building blocks for computer vision applications

Under active development, subject to API change

---

Pixano is an open-source tool by CEA List for exploring and annotating your dataset using AI features:

• Fast dataset navigation using the the modern storage format Lance
• Multi-view datasets support for text, images and videos, and soon for 3D point clouds
• Import and export support for dataset formats like COCO
• Semantic search using models like CLIP
• Smart segmentation using models like SAM

Installing Pixano

Production

We recommend installing the published Pixano package in a dedicated Python virtual environment (Python >= 3.10, < 3.14).

For example, with conda:

conda create -n pixano_env python=3.10
conda activate pixano_env

Then, install Pixano with pip:

pip install pixano

Init the database and start the Pixano server:

pixano init path/to/database
pixano server run path/to/database

Pixano is also available on the Docker Hub for official releases.

docker pull pixano/pixano:stable
docker run -p 7492:7492 -v /path/on/host:/library pixano/pixano:stable

In this example, /path/on/host is the directory on your machine where Pixano stores its library. Inside the container, Pixano always uses /library.

Development (from source)

To run the latest version of Pixano from source, you need uv. Install it if you haven't already:

curl -LsSf https://astral.sh/uv/install.sh | sh

Then, clone the repository and install all dependencies:

git clone https://github.com/pixano/pixano.git
cd pixano
uv sync

This installs the project in editable mode with all dependencies pinned via uv.lock. Your local changes are taken into account each time you run your environment.

Build the frontend UI assets:

cd ui/apps/pixano
pnpm i
pnpm run build

Init the data base and start the Pixano server:

uv run pixano init path/to/database
uv run pixano server run path/to/database

For more details on running Pixano locally (frontend setup, testing, formatting), see CONTRIBUTING.md.

Using Pixano

Please refer to our Getting started guide for information on how to launch and use the Pixano app, and how to create and use Pixano datasets.

Contributing

Please refer to our CONTRIBUTING.md for information on running Pixano locally and guidelines on how to publish your contributions.

License

Pixano is licensed under the CeCILL-C license.