Pixano
Palaiseau, France
What it is
Pixano is an open-source tool for exploring and annotating computer vision datasets. It works with multi-view collections containing text, images, and video, and provides annotation tools for boxes, polygons, pixel masks, keypoints, cuboids, classification, and tracking. Video labels can be propagated over time, while smart features include semantic search with models such as CLIP and segmentation with models such as SAM. Dataset formats such as COCO can be imported and exported, and Pixano uses Lance storage for dataset navigation. Developers can connect through REST and Python APIs, or assemble reusable Web Components into a custom application. Pixano Inference adds a Ray Serve-based inference server with a Python client and REST API for deployed models. Installation is documented through pip in a Python virtual environment or an official Docker image. The documented Python range is 3.10 or later and earlier than 3.14. Pixano is licensed under CeCILL-C and remains under active development, so its API may change.
For and against
For
- Annotation tools cover boxes, masks, cuboids, and tracking
- Supports text, image, and video datasets
- Imports and exports formats such as COCO
- Offers semantic search and model-assisted segmentation
- Provides REST and Python APIs
Against
- 3D point-cloud support is described as planned
- API may change during active development
- Requires Python 3.10 or later and earlier than 3.14
The verdict
Pixano combines dataset exploration, annotation, and AI-assisted tools with APIs and custom app components. Its active development status and planned point-cloud support are worth considering before adopting it.
Who it suits
Pixano is suited to AI developers and teams working with computer vision datasets, including applications in manufacturing, security, robotics, or transportation. It may suit users comfortable with documented Python or Docker installation.
On the record
- 3D annotation
- The official site says users can create cuboids and match them to point clouds using geometric transformations.pixano.cea.fr · read 30 Sep 2026
- AI features
- Pixano lists semantic search using models such as CLIP and smart segmentation using models such as SAM.github.com · read 30 Sep 2026
- Annotation tools
- The official site lists bounding boxes, editable polygons, pixelwise masks, customizable labels, and temporal propagation of video annotations.pixano.cea.fr · read 30 Sep 2026
- API
- Pixano documents a REST API and a Python API for interacting with the application and datasets.github.com · read 30 Sep 2026
- Custom apps
- Pixano's reusable annotation elements are Web Components that can be assembled into a custom app.pixano.cea.fr · read 30 Sep 2026
- Dataset formats
- It supports importing and exporting dataset formats such as COCO.github.com · read 30 Sep 2026
- Dataset navigation
- Pixano uses the Lance storage format for fast dataset navigation.github.com · read 30 Sep 2026
- Deployment
- The maker documents installation with pip in a Python virtual environment and official Docker releases.github.com · read 30 Sep 2026
- Development status
- The project states that it is under active development and subject to API changes.github.com · read 30 Sep 2026
- Import and export
- Pixano supports importing and exporting dataset formats such as COCO.github.com · read 30 Sep 2026
- Inference integration
- Pixano Inference provides a Ray Serve based inference server with a Python client and REST API for deployed models.github.com · read 30 Sep 2026
- Installation
- The project README describes installation with pip in a Python virtual environment or by running an official Docker image.github.com · read 30 Sep 2026
- Intended users
- CEA-List describes Pixano as supporting AI developers and applications in areas including manufacturing, security, robotics, and transportation.list.cea.fr · read 30 Sep 2026
- License
- Pixano is licensed under CeCILL-C.github.com · read 30 Sep 2026
- License and maturity
- Pixano is licensed under CeCILL-C, and its README says it is under active development and subject to API changes.github.com · read 30 Sep 2026
- Maker
- The product pages identify CEA List as Pixano's maker; its About page describes the LIST Institute as one of the three institutes of CEA Tech.pixano.cea.fr · read 30 Sep 2026
- Product
- Pixano is an open-source tool for exploring and annotating computer vision datasets with AI features.github.com · read 30 Sep 2026
- Purpose
- Pixano is an open-source tool for exploring and annotating computer vision datasets.pixano.github.io · read 30 Sep 2026
