Documentation
VisoLabel documentation
Everything you need to label data and train computer vision models — start to finish, all on your own machine. New here? Jump straight into the Quickstart.
The workflow: install → annotate → train → export → deploy
VisoLabel covers the whole computer vision loop in one local-first app. Each stage below links to its detailed guide.
Install
Download the native installer for Windows, macOS or Linux. No Docker, no drivers to wrangle.
Annotate
Label images and video with SAM2 assisted clicks and SAM3 bulk auto-annotation.
Train
Train RF-DETR detection, segmentation, and classification models on your GPU — or on Google Colab.
Export
Export weights and datasets to ONNX, PyTorch, or YOLO format with a single click.
Deploy
Run your trained model anywhere — on the edge, in the cloud, or back inside VisoLabel.
What is VisoLabel?
VisoLabel is a local-first computer vision platform. It bundles annotation and model training into a single desktop app so your data never has to leave your machine. You get assisted labeling powered by SAM2, bulk SAM3 auto-annotation, and one-click training of RF-DETR detection, segmentation, and classification models — on your own GPU or on Google Colab when you need more horsepower.
Unlike cloud annotation tools, there's no upload step, no per-seat data limits, and no third party touching your images. Everything from the first polygon to the exported model weights runs locally.
Explore the docs
Quickstart
→Go from a folder of raw images to a trained model in about ten minutes.
Annotation
→Master SAM2 assisted labeling, SAM3 auto-annotation, and video tracking.
Datasets & augmentation
→Review your gallery and multiply your data with preprocessing and augmentations.
Training
→Pick an RF-DETR model, set the split, and train locally or on Google Colab.
Keyboard shortcuts
→Label faster with the full list of annotation and navigation hotkeys.
Get help
Stuck on something? Check the FAQ for common questions, or browse the keyboard shortcuts to speed up your labeling. For anything else, reach out from the contact section on the homepage.