autodistill/autodistill
Autodistill automates the process of training small, fast supervised models from unlabeled images by leveraging large foundation models, eliminating the need for manual data labeling.
Core Features
Quick Start
pip install autodistillDetailed Introduction
Autodistill is an innovative open-source framework designed to streamline the creation of custom machine learning models, particularly for computer vision tasks. It addresses the critical bottleneck of data labeling by using powerful, albeit slower, foundation models to automatically label raw, unlabeled image data. This auto-labeled dataset then trains smaller, faster supervised models, which are ideal for efficient deployment on edge devices or in the cloud. The framework provides a pluggable architecture, allowing users to define tasks, select base models, and specify ontologies to generate highly specialized and performant distilled models without human intervention.