Uniform Recognizer

The Uniform Recognizer detects US military uniform decorations using a trained neural network object detector model.

The training pipeline and model are made available to the public.

The trained model can be tried out as an add-on of the free NAVADMIN Viewer iOS app.

Training pipeline

The training pipeline consists of:

• Processing image files renameAndStripEXIFEachDirectory.sh

• Annotating ground truths in sloth

  • Run ../run_sloth.sh inside each image directory

• Downscaling image files downscaleAnnotationsAndLinkedImages.py

• Training an object detection model in turicreate

  • Run trainTheMachine.py -t

• Evaluate and export model evaluateModel.py

• See trainTheMachine/training-pictures for directions.

Models

• Trained models will be distributed as releases.
• See models for released models.

Acknowledgements

• cvhciKIT/sloth annotation tool

• apple/turicreate machine learning models

• PBS Professional job scheduling

• This work was supported in part by a grant of computer time from the DoD High Performance Computing Modernization Program at Navy DSRC, ERDC, and ORS.

License

• All original work in this repository is released under LGPL v3

  • Attribution to Anson Liu required for end products utilizing models created from this repository.

• All dependencies retain their original licenses.