Launch: Model Prompting for Automated Labeling with Autodistill

Roboflow has significantly advanced automated image labeling by introducing Auto Label, a fully hosted product that leverages model-assisted annotation and over 50,000 trained models. Their Smart Polygon tool, powered by the Segment Anything model, enhances labeling efficiency with polygon-based annotations. In June 2023, Roboflow launched Autodistill, an open-source framework that integrates foundation models such as Grounding DINO, CLIP, and Segment Anything to facilitate auto-labeling for diverse applications. This tool allows users to assess how multimodal foundation models perform on specific datasets, streamlining the traditionally labor-intensive labeling process. Users can configure Autodistill for labeling tasks through a straightforward process involving creating a project, uploading data, and utilizing an interface to test and evaluate model performance. Roboflow also provides solutions for those needing additional support through outsourced labeling services.