ReceiptMoa is built on the open source software, recognition models, fonts and public data listed below. None of these files are loaded from third-party servers; they're all served from the same address as the service.
Text recognition
| Name | Used for | License |
|---|---|---|
| PaddleOCR PP-OCRv5 mobile models (PP-OCRv5_mobile_det, korean_PP-OCRv5_mobile_rec, en_PP-OCRv5_mobile_rec) | Finding text on receipt photos and reading it | Apache License 2.0 (notice) |
| ONNX Runtime Web | Running the recognition models in the browser | MIT License |
| Tesseract.js and tesseract.js-core | A fallback engine used if the models can't be loaded | Apache License 2.0 |
| Tesseract language data (Korean, English) | Language data for the fallback engine | Apache License 2.0 |
Photos and files
| Name | Used for | License |
|---|---|---|
| libheif (libheif-js) | Converting iPhone HEIC photos in browsers that can't read them | LGPL-3.0 (full license) |
| SheetJS Community Edition | Creating the Excel (.xlsx) file | Apache License 2.0 |
libheif-js is used unmodified, loaded as a separate file and only downloaded when needed.
Fonts
| Name | License |
|---|---|
| SUIT Variable | SIL Open Font License 1.1 (full license) |
Public data
| Name | Used for | Terms |
|---|---|---|
| Small Enterprise and Market Service commercial district data, Korea Public Data Portal | Building a nationwide store name list to confirm Korean store names | No usage restrictions |
The store list is downloaded whole, regardless of what's on your receipts, and compared only inside your browser. It contains store names only, and nothing from your receipts is sent anywhere.
Contact
If a credit is missing or needs correcting, please let us know at the email address on the Contact page.