app model files: README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: he
|
| 3 |
+
tags: [ocr, handwriting, hebrew, onnx, android]
|
| 4 |
+
---
|
| 5 |
+
# Hebrew handwriting OCR: on-device model files
|
| 6 |
+
|
| 7 |
+
Files downloaded by the Android app (github.com/cyttic/android-hwrg), which runs the whole pipeline on the phone.
|
| 8 |
+
|
| 9 |
+
| File | What |
|
| 10 |
+
|---|---|
|
| 11 |
+
| `line_int8.onnx` | CTC line model `cyttic/exp35-trocr-base-en-ctc-seqhead-allreal-synpunct` (TrOCR-base ViT encoder + 2-layer transformer CTC head, 198 classes), ONNX opset 17, dynamic int8 (all constant-weight MatMuls except the ViT layer-2 MLP fc2, kept in fp32). Inputs: `pixel_values` (B, 3, 384, 384) in [-1, 1], `cols` (B,) int64; output `log_probs` (B, 288, 198). |
|
| 12 |
+
| `charset.json` | `chars[k]` = character of class k; `chars[0]` = CTC blank. |
|
| 13 |
+
| `heb_char6_kn.bin.gz` | Character 6-gram LM, interpolated Kneser-Ney (Ben-Yehuda + DictaGen text), binary layout in `tools/export_lm.py`. |
|
| 14 |
+
|
| 15 |
+
The line finder is `PaddlePaddle/PP-OCRv5_server_det_onnx` (Apache-2.0), downloaded from its own repo.
|