oneopen-annotator 0.1.0

Open-source localhost data preparation and training framework with YOLO & SAM auto-annotation

OneOpen Annotator

Open-source, localhost-only data preparation & training
Annotate, generate versions, export, and train YOLO — all on your machine.

---

Install

pip install oneopen-annotator

With extras:

# YOLO Label Assist + local training (Ultralytics)
pip install "oneopen-annotator[yolo]"

# LibreYOLO backend
pip install "oneopen-annotator[libreyolo]"

# SAM interactive segmentation
pip install "oneopen-annotator[sam]"

# Everything
pip install "oneopen-annotator[all]"

From source (editable):

git clone https://github.com/oneopen/oneopen-annotator.git
cd oneopen-annotator
pip install -e ".[yolo]"

Quick start

oneopen init          # create ~/.oneopen
oneopen serve         # http://127.0.0.1:8765

Or:

python -m oneopen_annotator serve

Open http://127.0.0.1:8765 — no account, no cloud upload.

What it does

Step	Capability
Projects	Detection, segmentation, classification — create or import folder/zip
Upload	Bulk images, stored under your chosen project location
Annotate	Boxes & polygons; classes required before labeling
Label Assist	YOLO auto-label (YOLOv8–v10, YOLO11, YOLO26, LibreYOLO, custom weights)
SAM	Interactive segmentation (optional [sam])
Generate	Preprocess, augment, train/valid/test splits, dataset versions
Train / Export	Local YOLO training; export YOLO zip + COCO

100% local: SQLite + files under ~/.oneopen (or a path you pick per project).

CLI

oneopen serve                 Start the web UI (default :8765)
oneopen serve --port 9000     Custom port
oneopen init                  Initialize local data directory
oneopen version               Print version

Environment:

# optional — override data root
set ONEOPEN_DATA_DIR=D:\oneopen-data   # Windows
export ONEOPEN_DATA_DIR=/data/oneopen # Unix

Data layout

Path	Contents
~/.oneopen/oneopen.db	SQLite database
~/.oneopen/projects/<id>/	Default project location (images/)
~/.oneopen/exports/	Exported datasets
~/.oneopen/models/	Weights & training runs

Optional: SAM

1. Install: pip install "oneopen-annotator[sam]" and segment-anything.
2. Place a checkpoint at ~/.oneopen/models/sam_vit_b_01ec64.pth (or configure equivalently).

Requirements

• Python 3.10+
• A modern browser
• Optional GPU for faster YOLO/SAM/training

License & credits

Apache License 2.0 — free to use, modify, and redistribute, including commercially.

Redistributors must keep copyright notices and the NOTICE file, so author credit stays with the project.

Copyright 2026 oneopensource

See LICENSE and NOTICE.

Citation

If you use OneOpen Annotator in research or a product, please credit:

@software{oneopensource_oneopen_annotator_2026,
  author = {oneopensource},
  title  = {OneOpen Annotator},
  year   = {2026},
  url    = {https://github.com/oneopen/oneopen-annotator},
  note   = {Open-source localhost CV annotation and training}
}

Author

oneopensource — creator and maintainer of OneOpen Annotator.

Contributions welcome under the same Apache 2.0 license.