Repolode
Submit
xtreme1-io

xtreme1-io/xtreme1

Open-source all-in-one platform for annotating and curating multimodal training data, covering images, 3D LiDAR point clouds, and LLM RLHF.

Open-source alternative to:Scale AI

View on GitHubWebsite
Stars
1,385
Forks
222
Open issues
22
Last commit
24 days ago

What is it?

Xtreme1 is a self-hosted data labeling and annotation platform for building machine learning training datasets. It covers 2D image annotation (bounding boxes, polygons, polylines, key points), 3D LiDAR point cloud labeling, and LiDAR-camera sensor fusion, aimed at tasks such as 2D/3D object detection and 3D semantic or instance segmentation.

Beyond labeling, the platform includes data management and quality monitoring, tools for finding and fixing labeling errors, a configurable ontology for classes with hierarchies and attributes, and visualization of model results. Built-in pre-labeling models run in containers and need an NVIDIA GPU. An RLHF annotation tool for large language models is available in beta.

The stack runs with Docker Compose, using MySQL, Redis and MinIO for storage, and the release package includes a sample dataset. The source can be built locally for those who want to extend it. Releases are Apache 2.0 licensed and hosted under LF AI & Data Foundation.

Key features

  • 01Annotates 2D images with bounding boxes, polygons, polylines and key points.
  • 02Labels 3D LiDAR point clouds and supports LiDAR-camera sensor fusion.
  • 03Provides built-in pre-labeling models for 2D/3D object detection.
  • 04Offers a configurable ontology with class hierarchies and attributes.
  • 05Includes data management, quality monitoring and labeling error detection.
  • 06Includes a beta RLHF annotation tool for large language models.

Use cases

  • Labeling camera and LiDAR data for autonomous driving or robotics perception models.
  • Building 2D or 3D object detection and segmentation datasets.
  • Curating and visualizing image datasets to debug model training data.
  • Creating preference or feedback data for RLHF on large language models.

Quick start

  1. 1

    Download the latest release package

    wget https://github.com/xtreme1-io/xtreme1/releases/download/v0.9.3/xtreme1-v0.9.3.zip
  2. 2

    Unzip the release package

    unzip -d xtreme1-v0.9.3 xtreme1-v0.9.3.zip
  3. 3

    Enter the package directory and start all services with Docker Compose

    docker compose up
  4. 4

    Open http://localhost:8190 in your browser (Google Chrome recommended)

Things to know

  • Requires Docker, and the first start takes a few minutes to initialize the database.
  • Built-in model containers only run on Linux servers with an NVIDIA GPU.
  • The RLHF for LLM feature is in beta.

In collections

Computer vision libraries & toolsReinforcement learning frameworks & training environments

Topics

#data-labeling#annotation#computer-vision#point-cloud#lidar#multimodal#object-detection#rlhf#3d-annotation#annotation-tool

FAQ

What is xtreme1?
Open-source all-in-one platform for annotating and curating multimodal training data, covering images, 3D LiDAR point clouds, and LLM RLHF. Xtreme1 is a self-hosted data labeling and annotation platform for building machine learning training datasets. It covers 2D image annotation (bounding boxes, polygons, polylines, key points), 3D LiDAR point cloud labeling, and LiDAR-camera sensor fusion, aimed at tasks such as 2D/3D object detection and 3D semantic or instance segmentation.
Is xtreme1 free to use? What license does it have?
xtreme1 is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is xtreme1 still maintained?
Actively maintained. The last commit to xtreme1 was on Sep 29, 2026.
What can I use xtreme1 for?
Labeling camera and LiDAR data for autonomous driving or robotics perception models. Building 2D or 3D object detection and segmentation datasets. Curating and visualizing image datasets to debug model training data. Creating preference or feedback data for RLHF on large language models.
How do I install xtreme1?
Download the latest release package: wget https://github.com/xtreme1-io/xtreme1/releases/download/v0.9.3/xtreme1-v0.9.3.zip Unzip the release package: unzip -d xtreme1-v0.9.3 xtreme1-v0.9.3.zip Enter the package directory and start all services with Docker Compose: docker compose up Open http://localhost:8190 in your browser (Google Chrome recommended)
Where can I get xtreme1?
The source code and installation instructions are on GitHub: https://github.com/xtreme1-io/xtreme1

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects