| --- |
| license: apache-2.0 |
| library_name: onnx |
| tags: |
| - ocr |
| - handwriting-recognition |
| - paddleocr |
| - onnx |
| - multilingual |
| language: |
| - en |
| - zh |
| - ja |
| - hi |
| - th |
| - ar |
| - ko |
| --- |
| |
| # DocScanner on-device OCR models |
|
|
| ONNX exports of [PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR) text detection and |
| recognition models, used by [DocScanner](https://github.com/Bible-Translation-Tools/doc-scanner) |
| to transcribe handwritten pages **entirely on the device**. |
|
|
| DocScanner is a free, open-source tool for Bible translation field teams. Transcription has to |
| work without a network and without per-page cost, so no cloud model is involved: a page is cut |
| into text lines by the detector, and each line is read by the recognizer for that project's |
| writing system. |
|
|
| ## Layout |
|
|
| A device downloads the detector plus exactly one recognizer — 13 MB for most scripts, 78 MB for |
| Latin/CJK — never the whole set. |
|
|
| | path | file | size | sha256 (short) | |
| |------|------|------|----------------| |
| | `detector/` | `det_model.onnx` | 4.6 MB | `0c5eeee2` | |
| | `ppocrv6/` | `rec_model.onnx` | 73 MB | `4078550d` | |
| | `ppocrv6/` | `charset.json` | 128 KB | `46f80089` | |
| | `devanagari/` | `rec_model.onnx` | 7.6 MB | `a3d5b5fa` | |
| | `devanagari/` | `charset.json` | 3.6 KB | `621d0073` | |
| | `thai/` | `rec_model.onnx` | 7.5 MB | `d52231fe` | |
| | `thai/` | `charset.json` | 3.3 KB | `b62ede69` | |
| | `arabic/` | `rec_model.onnx` | 7.6 MB | `a1e69c68` | |
| | `arabic/` | `charset.json` | 4.5 KB | `df730929` | |
| | `korean/` | `rec_model.onnx` | 13 MB | `03525a1d` | |
| | `korean/` | `charset.json` | 81 KB | `e600744e` | |
|
|
| The app pins a **tag** of this repository in its download URL, so a given build can only ever |
| fetch the weights it was tested against. |
|
|
| ## Provenance |
|
|
| Exported with [`paddle2onnx`](https://github.com/PaddlePaddle/Paddle2ONNX) from the official |
| PaddleOCR inference models, unquantized (float32): |
|
|
| | directory | source model | opset | |
| |-----------|--------------|-------| |
| | `detector/` | `PP-OCRv5_mobile_det` | 17 | |
| | `ppocrv6/` | `PP-OCRv6_medium_rec` | 16 | |
| | `devanagari/` | `devanagari_PP-OCRv5_mobile_rec` | 16 | |
| | `thai/` | `th_PP-OCRv5_mobile_rec` | 16 | |
| | `arabic/` | `arabic_PP-OCRv5_mobile_rec` | 16 | |
| | `korean/` | `korean_PP-OCRv5_mobile_rec` | 16 | |
|
|
| `charset.json` is each model's character table from PaddleOCR, as a JSON array. The recognizer |
| output has `charset + 2` classes: index 0 is the CTC blank, then the table, then a space — |
| the layout PaddleOCR's `CTCLabelDecode` expects. |
|
|
| ## Interfaces |
|
|
| **Detector** (`det_model.onnx`) — input `[1, 3, H, W]`, BGR, long side scaled to 960 and floored |
| to a multiple of 32, normalized with ImageNet statistics (mean `0.485/0.456/0.406`, std |
| `0.229/0.224/0.225`). Output `[1, 1, H, W]`, a per-pixel probability of text. |
|
|
| **Recognizers** (`rec_model.onnx`) — input `[1, 3, 48, W]`, RGB, height 48 with width scaled by |
| the line's aspect ratio and padded to a multiple of 32, normalized `(x/255 - 0.5) / 0.5`. |
| Padding is left at zero *after* normalization (mid-grey), which is how PaddleOCR pads; padding |
| with black instead wrecks recognition. Output `[1, T, classes]` logits, greedy CTC decoded. |
|
|
| Arabic is read right-to-left: CTC scans left to right and so emits the logically last character |
| first. The line is reversed by grapheme cluster, keeping combining marks attached to their base |
| letter — correcting this took the error on rendered Arabic from 79% to 18%. |
|
|
| ## Measured accuracy |
|
|
| Character error rate on handwritten sample pages, per line, after detection: |
|
|
| | script | CER | note | |
| |--------|-----|------| |
| | English (neat) | 5-7% | | |
| | English (hard hand) | ~14% | | |
| | Spanish / French | ~4% | | |
| | Chinese | 2% | written in a squared-paper grid | |
| | Devanagari | usable | whole lines, most words legible | |
| | Thai | 15-25% | | |
| | Arabic | 25-35% | words land in the right places and reading order | |
| | Korean | 15-25% | several lines nearly verbatim | |
|
|
| These are CTC models with **no language model**, which is deliberate: an unreadable crop comes |
| back garbled or empty rather than as fluent invented text. A TrOCR alternative scored better on |
| neat English (2%) but rewrote what it could not read — including inventing liturgical Church |
| Slavonic from a blank strip — which is the wrong failure mode for a translation tool. |
|
|
| ## Not included |
|
|
| - **Cyrillic** — PaddleOCR's Cyrillic recognizers manage only 76-84% CER on handwriting, so |
| those projects use server transcription until a model trained on handwritten lines exists. |
| - **Hebrew** — no PaddleOCR model. |
| - **Greek** — not yet wired. |
|
|
| ## License |
|
|
| Apache-2.0, inherited from PaddleOCR. Please keep the attribution to the PaddleOCR project. |
|
|
