OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Autodistill

Autodistill is an open-source Python library and plugin framework for generating image labels with foundation models and using the resulting datasets to train supervised computer-vision models.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Autodistill — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Autodistill when you want to connect foundation-model-assisted image labeling to training a smaller supervised vision model, and can select compatible model plugins for your task.

Who it is for

  • Python developers building image-labeling and computer-vision training workflows.
  • Teams with image collections that want to create annotations using foundation models before training a task-specific model.

Documented uses

  • Generate image annotations from prompts or other ontology definitions.
  • Create a labeled dataset for supervised model training.
  • Train a target model and use it to produce predictions on images.
Before you choose

Requirements & limits

The public GitHub repository is listed under the Apache-2.0 license. The README describes running the library on local hardware and separately mentions a Roboflow-hosted labeling option.

  • The library package alone does not implement every model: compatible Base Model and Target Model plugins must also be installed.
  • Autodistill’s ontology defines how prompts map to dataset labels, and task compatibility is required across pipeline components.
  • The project describes trained target models as smaller and faster than foundation models, with weaker generalization beyond their training dataset.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Create a labeled dataset and train a YOLOv8 model

Use a foundation model to label images and train a supervised detection model through Autodistill’s documented command-line workflow.

Before you start

  • Python environment with the autodistill, autodistill-grounded-sam, and autodistill-yolov8 packages installed.
  • An images directory containing the input images.
  • A prompt and label to define the ontology.

Steps

  1. Install the core library and the GroundedSAM and YOLOv8 plugins.

  2. Run the documented autodistill images command with Grounding DINO as the base model, YOLOv8 as the target, an ontology prompt and label, and an output directory.

  3. Use the generated dataset as the training input for the selected target model.

Expected outcome

The README says the command labels the images and saves the resulting dataset in the specified output folder; the workflow uses that dataset to train a YOLOv8 model.

Guide limitations

  • The task and ontology must match across the base model, ontology, and target model.
  • Target models are described as having limited generalization beyond information represented in their dataset.
  • The README’s command-line example uses Grounding DINO, while its package-install example names GroundedSAM; check the model-specific plugin instructions for the selected combination.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
10 October 2026
Original website title
autodistill/autodistill · Official homepage title · retained in its source language
How we review sourcesSuggest a correction