Image Annotation Tool

The tool provides a structured image annotation workflow for developers, data scientists, and ML engineers. It supports multiple annotation types (bounding boxes, polygons, and keypoints) and allows labeling with custom classes, attributes, and confidence scores.
Conceptually, users provide an image source (URL or file), choose an annotation type, and draw shapes over the image. Each shape stores coordinates in pixels and maps to a class label. Annotations are stored in a project-scoped schema and can be validated before export.
Outputs include machine-readable datasets in COCO JSON, YOLO TXT, and Pascal VOC XML formats, plus a compact per-annotation JSON. Exports preserve image metadata, class mappings, and coordinate formats for downstream training.
Use cases span object detection, instance segmentation, and dataset curation for GitHub-based ML projects. It enables reproducible labeling pipelines, supports import of existing annotations, and integrates with version control to fit standard ML workflows.