LabelU
github.com
- Free plan
- Windows
- macOS
- Web
- API
- Self-hosted
What LabelU is
LabelU is an open-source platform for annotating image, video and audio data, aimed at work supporting data analysis and model training. Image tools include bounding boxes, semantic segmentation, polylines and keypoints; video and audio tools handle segmentation, classification and information extraction. Users can load pre-annotated data and refine it, or use AI model services to detect and segment image objects, including batch jobs with progress tracking. The platform can import annotation data from S3-compatible storage such as AWS S3 and MinIO, then export JSON, COCO and MASK formats. Local deployment uses Miniconda, Python 3.11 and pip, with a server at localhost:8000. LabelU includes SQLite and supports MySQL installation and migration. Its model server exposes HTTP endpoints, including POST / and GET /health. Reference model requirements vary: Florence-2 and GroundingDINO with SAM ViT-B each require about 4GB VRAM, while SAM 3 requires about 8GB and CUDA 12.6 or later. The project is released under Apache 2.0. The README invites users to join the OpenDataLab WeChat group for support.
Who it suits
LabelU suits teams annotating image, video or audio datasets, including workflows that combine model-assisted labels with human review. Its self-hosted deployment options may appeal to groups able to manage the listed local setup and model hardware.
Good to know, and what to watch
- Supports image, video and audio annotation.
- Can refine pre-annotated data.
- Imports from AWS S3 and MinIO.
- Exports JSON, COCO and MASK.
- Includes a unified model-server HTTP API.
- Local setup requires Miniconda and Python 3.11.
- SAM 3 requires about 8GB VRAM and CUDA 12.6+.
- Support channel listed is an OpenDataLab WeChat group.
The verdict
LabelU combines multimodal annotation tools with model-assisted labeling, flexible exports and self-hosted deployment. Check the setup and model hardware requirements against the environment available to your team.
What the maker publishes
- AI assistance
- Users can load pre-annotated data in one click and refine it in the platform. source
- AI assisted labeling
- Users can load pre-annotated data with one click and refine or adjust it. source
- AI auto-annotation
- AI model services can automatically detect and segment image objects, including batch annotation with real-time progress tracking. source
- AI auto-annotation
- Image auto-annotation can detect and segment objects, with batch task annotation and real-time progress tracking. source
- API
- The model server exposes a unified HTTP API with POST / and GET /health endpoints. source
- Audio annotation
- Audio tools support audio segmentation, audio classification, and audio information extraction. source
- Audio tools
- Audio annotation supports segmentation, classification, and information extraction. source
- Database support
- LabelU includes built-in SQLite and supports MySQL installation and migration. source
- Database support
- SQLite is the documented default database, and MySQL can be installed as an optional dependency. source
- Deployment
- Local deployment uses Miniconda, Python 3.11, pip installation, and a local server at http://localhost:8000/. source
- Deployment
- The project documents local installation with Python 3.11 or later and running a local server at localhost:8000. source
- Export formats
- The platform supports exporting data in JSON, COCO, and MASK formats. source
- Export formats
- The README lists JSON, COCO, and MASK among supported export formats. source
- Image annotation
- Image tools include 2D bounding boxes, semantic segmentation, polylines, and keypoints. source
- Image tools
- Image annotations include 2D bounding boxes, semantic segmentation, polylines, and keypoints. source
- Intended users
- The README describes the platform as suited to annotation work supporting complex data analysis and model training. source
- License
- The project is released under the Apache 2.0 license. source
- License
- The project is released under the Apache 2.0 license. source
- Model requirements
- Florence-2 requires about 4GB VRAM, GroundingDINO plus SAM ViT-B about 4GB, and SAM 3 about 8GB with CUDA 12.6+. source
- Object storage
- LabelU can import annotation data from S3-compatible storage such as AWS S3 and MinIO. source
- Purpose
- LabelU is an open-source multimodal data annotation platform for image, video, and audio data. source
- Purpose
- LabelU is an open-source multimodal data annotation platform for image, video, and audio tasks. source
- Reference models
- Reference model servers include Florence-2, GroundingDINO plus SAM ViT-B, and SAM 3. source
- Reference models
- The project lists Florence-2, GroundingDINO with SAM ViT-B, and SAM 3 reference model servers, with GPU memory requirements stated for each. source
Checked against the maker's pages on 2 Oct 2026.