Pixano

Overview

Pixano is an open-source workspace for exploring and labeling computer-vision datasets, aimed especially at AI developers who want to shape annotation into their own workflows.

Made by CEA List and licensed under CeCILL-C, Pixano supports datasets combining text, images and video, with COCO among its import and export formats. Teams can install it with pip in a Python virtual environment or use an official Docker image; its documented Python range is 3.10 or later and earlier than 3.14. The project is under active development, so teams building integrations should be prepared for API changes.

Key features

Annotation and AI assistance

Pixano spans bounding boxes, polygons, pixel masks, keypoints, cuboids, classification and tracking. Custom labels and propagation of video annotations over time extend it beyond static image labeling. For 3D work, users can create cuboids and align them with point clouds through geometric transformations. That range suits teams with varied vision tasks; point-cloud support within multi-view datasets is described as planned, so it is not a reason to choose Pixano for a mature point-cloud workflow today.

Model-assisted labeling includes smart segmentation with models such as SAM, while semantic search can use models such as CLIP to find relevant dataset content. These capabilities can help teams organize and label large collections, especially when they already use those model families. Pixano does not present itself as a turnkey labeling service, so teams seeking a managed end-to-end operation may prefer another tool.

Dataset workflows and extensibility

Pixano uses Lance for dataset navigation and storage, and supports formats such as COCO for import and export. Annotation exports include COCO and Pixano's own format; JPEG and PNG are supported image formats. REST and Python APIs provide ways to interact with the application and datasets. Reusable annotation elements are Web Components that developers can assemble into a custom app, making Pixano particularly attractive when a team wants to build a tailored interface rather than adopt a fixed one.

Pixano Inference adds a Ray Serve-based inference server, with a Python client and REST API for deployed models. That gives model developers an integration point alongside annotation and dataset tools, though the project's stated active development and possible API changes call for caution before making it a critical, long-lived dependency.

Pricing

Pixano is free, with a free plan and no free trial. Its open-source licensing makes it a viable option for teams that can deploy and manage the software themselves rather than pay for a hosted subscription. There are no paid tiers or published seat or usage quotas to weigh against a cheaper plan.

The trade-off is operational: installation is through pip or Docker, and the documented setup expects Python 3.10 or later but earlier than 3.14. Teams should also account for integration upkeep as the project evolves. Pixano's pages do not state a security certification or compliance standard, so organizations with formal requirements should assess suitability before adoption.

Platforms

Pixano is available through web, API, self-hosted, Linux, macOS and Windows options. Its self-hosted deployment paths suit teams that want to run the software in their own environment, while the APIs and Web Components serve developers building around it. The Python version range is a practical constraint for deployment planning.

Who it's for

Pixano is best suited to AI developers and application teams working with computer-vision datasets, including the manufacturing, security, robotics and transportation domains identified by CEA-List. It is a strong fit for teams that need both annotation breadth and room to customize their application. It is less suitable for buyers prioritizing a settled API, a managed service, or documented compliance assurances.

Pros and cons

• Pros: Broad annotation coverage, from 2D shapes and masks to tracking and cuboids, supports varied computer-vision tasks in one tool.
• Pros: CLIP-based semantic search, SAM-style smart segmentation and video-label propagation add useful assistance beyond manual annotation.
• Pros: Free access, COCO exchange, REST and Python APIs, and reusable Web Components give technically capable teams room to build their own workflow.
• Cons: Active development and possible API changes create maintenance risk for teams depending on stable integrations.
• Cons: The Python version requirement and pip or Docker deployment call for technical ownership rather than a plug-and-play setup.
• Cons: Point-cloud support for multi-view datasets is planned, not established, and no security certification or compliance standard is stated.

Alternatives

Label Studio is worth considering for teams seeking another free-plan option with a free trial and support across web, self-hosted, desktop and API environments; its Community Edition is its named plan.

Roboflow may suit teams that want a freemium platform with Android and iOS as well as web and self-hosted access. Its Free Tier includes 10 credits a month, described as enough to train about 30 models or run 80,000 inferences; its Core plan is 39.00 USD per month.

BasicAI is an option for buyers considering private-cloud deployment, with a plan starting at 6600.00 USD per year and customizable payment cycle, seats, storage, model calls and on-premise deployment options.

CVAT offers a free Community plan for personal use and small teams under an MIT license, as well as a free online plan with stated member, project, task and storage limits. Labelbox is another freemium, free-plan alternative available through web and API.

MakeSense.ai is a free alternative available on web, Windows, macOS and Linux. Ultralytics Platform offers a freemium plan with published storage, upload, model, training-job and deployment allowances for readers comparing fixed cloud limits. Labelme is another freemium option for Windows, macOS and Linux.

Readers comparing more tools can browse Image Annotation Software and AI Image Annotation Tools.

Verdict

Choose Pixano if you are an AI developer or computer-vision team that wants free, extensible dataset exploration and annotation, with model-assisted search and labeling in the same toolkit. Its breadth and customizable components are the strongest reasons to pick it. Look elsewhere if your workflow needs stable APIs, established point-cloud support, or a documented security or compliance posture.