coco-annotator

Why we included this project

Building a detection or segmentation dataset usually means either writing your own annotation tool or paying for a hosted service. This web app gives you a self-hosted alternative: a browser-based editor where a small team can draw polygons and keypoints, and track object instances across frames, all from a normal workstation. You run it locally or through the provided Docker image, and the server manages users and projects, so multiple annotators can work on the same dataset without stepping on each other. Export goes straight to the COCO schema, the format most detection pipelines expect, so you skip the error-prone conversion step. It's a good fit for teams that want control over their labeling workflow.