athewsey
commented
3 years ago - Improve Rek model start-up triggering behaviour in the pre-processing Lambda function
- Add a bucket to the OCR stack for Rek training data (with permissions pre-configured to simplify setup)
- Update docs for new SMGT "Automated setup" workflow (replaces old "Create manifest file" hyperlink in console for create labelling job) and general improvements
- Remove old receipt images dataset from source code (preprocessing README now links to S3 hosted custom dataset)