AnnoMage: A Semi-Automatic Image Annotation Tool

abstract

Anno-Mage is a semi-automatic image annotation tool that uses a RetinaNet ResNet50 FPN V2 (PyTorch) model to suggest bounding box annotations, which users can then confirm, correct, or supplement manually. It significantly reduces the time and effort required to build annotated datasets for object detection tasks. Annotations are exported as CSV or Pascal VOC XML.

demo

Download Source (GitHub)

user guide

installation

Install from PyPI:

pip install anno-mage

Then launch:

anno-mage

instructions

1. Launch the tool: anno-mage
2. Select the folder of images to annotate.
3. Enter your object class names.
4. RetinaNet ResNet50 FPN V2 runs automatically and suggests bounding boxes for each image.
5. Refine boxes via drag-and-drop; navigate images with arrow keys — annotations save automatically.

usage

Annotations are written to the annotations/ directory inside your image folder:

• CSV: annotations/annotations.csv — columns: image_path,x1,y1,x2,y2,label
• Pascal VOC XML: annotations/annotations_voc/ — one .xml file per image

Custom model support: legacy Keras .h5 weights and TensorFlow frozen_inference_graph.pb files can be loaded in place of the default PyTorch model.

community

Join the discussion on Slack.

acknowledgments

1. Meditab Software Inc. — for supporting this project.
2. PyTorch / Torchvision teams — the object detection backbone powering the auto-suggestion engine.
3. Computer Vision Group, LDCE — for feedback and testing.