lost

Why we included this project

LOST is a web-based image annotation framework for teams that have outgrown single-user labeling tools. You run it in the browser, distribute labeling tasks to annotators anywhere, and mark up bounding boxes, polygons, points, or lines with the built-in tools. What makes it interesting is the semi-automatic angle: you can feed AI-generated proposals into the workflow so annotators review and correct machine suggestions instead of drawing every object by hand. The ready-made pipelines work without any programming, but you can also build your own by mixing annotation interfaces with custom algorithms, and the Jupyter integration keeps that from getting too technical. On the practical side, it connects to S3, Azure blob storage, or FTP, organizes labels into colored trees, tracks progress, and exports annotations whenever you need them.