Diffgram
lastmod 2026-11-10created 2022-05-112 edits
Diffgram is an open-source annotation platform for creating training data for computer vision models. Along with a web UI for attaching labels such as bounding boxes, polygons, and keypoints to images and videos, it bundles features needed when several people work on labeling together, such as task distribution, review, and dataset versioning. It can be installed and run on your own server, and it connects to training pipelines through a Python SDK.
Open Source Image Annotation. Accelerate Your Computer Vision AI Model Training