Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

Xtreme1

github.com

  • Free plan
  • Windows
  • macOS
  • Linux
  • Web
  • Self-hosted
7.8 Club score Visit github.com

What Xtreme1 is

Xtreme1 is an open-source platform for annotating, curating, and organizing multimodal training data. It supports images, 3D LiDAR, and 2D/3D sensor-fusion datasets, with tools for object detection, 3D segmentation, LiDAR-camera fusion, data-quality monitoring, and model-result visualization. Annotation types include bounding boxes, polygons, polylines, and classification, with image formats including JPG and PNG. Built-in pre-labeling and interactive models assist detection, segmentation, and classification. The configurable Ontology Center organizes training classes with hierarchies and attributes. Image workflows link with YOLOR and RITM, while LiDAR-camera workflows link with OpenPCDet and AB3DMOT. RLHF annotation for large language models is listed as a beta feature. Docker Compose deployment runs on Linux, Windows, or MacOS, and the project offers API access and exports to COCO, JSON, and Xtreme1 formats. BasicAI is identified as the cloud version; private-cloud deployment starts from $6,600/y.

Who it suits

Xtreme1 suits machine-learning teams preparing image, LiDAR, sensor-fusion, or language-model datasets. It can serve self-hosted groups needing ontology management, model assistance, and API-based workflows.

Good to know, and what to watch

  • Open-source Apache 2.0 software
  • Supports image and 3D sensor datasets
  • Includes model-assisted labeling
  • Exports COCO, JSON, and Xtreme1 formats
  • Runs on Linux, Windows, and MacOS
  • RLHF annotation is still beta
  • Built-in models require NVIDIA CUDA and T4-class GPU
  • Private-cloud deployment starts from $6,600/y
  • Requires at least 10GB free disk space

Private-Cloud Deployment

$6,600/yr

Open source

The verdict

Xtreme1 provides broad annotation and ontology tooling for multimodal training data, with self-hosted deployment and model-assisted labeling. Teams should account for GPU requirements when using built-in model containers and for the separate cost of private-cloud deployment.

What the maker publishes

Annotation features
Its listed capabilities include 2D/3D object detection, 3D segmentation, LiDAR-camera fusion, data quality monitoring and model-result visualization. source
Cloud version
The documentation identifies BasicAI as the cloud version and the GitHub README links to a free enterprise demo request. source
Data types
The project says it supports labeling images, 3D LiDAR and 2D/3D sensor-fusion datasets. source
Deployment
The open-source project is installed with Docker Compose and can be run on Linux, Windows or MacOS. source
Integrations
The feature page links image annotation with YOLOR and RITM, and LiDAR-camera fusion workflows with OpenPCDet and AB3DMOT. source
License
The software is licensed under Apache 2.0, and the README identifies Xtreme1 as an LF AI & Data Foundation project. source
LLM support
The GitHub README lists RLHF annotation for large language models as a beta feature. source
Ontology
The configurable Ontology Center supports classes with hierarchies and attributes for model training. source
Pre-labeling
Built-in pre-labeling and interactive models support 2D/3D object detection, segmentation and classification. source
Pricing limits
The BasicAI platform pricing page lists private-cloud deployment starting from $6,600/y, with seats, storage and model calls customizable. source
Requirements
The README lists AMD64 or ARM64 CPUs, at least 2GB RAM and 10GB or more free disk space; built-in model containers require a Linux server with NVIDIA CUDA and a T4 or similar GPU. source
Security reporting
The project security policy says it follows the Linux Foundation process for reporting vulnerabilities to LF projects and foundations. source
What it does
Xtreme1 is an open-source platform for multimodal training data that supports data annotation, curation and ontology management. source

Checked against the maker's pages on 9 Oct 2026.