Instructions to use lastabhi/amar-line-readers with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LiteRT
How to use lastabhi/amar-line-readers with LiteRT:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Amar line readers
Models that let the Amar Vault Android app read text from screenshots and scanned pages line by line, entirely on the phone. A finder marks every line of text in a picture, then each line is read by a small reader: one for all text, one for lines with numbers, one for Hindi (Devanagari).
The app tries this way of reading against its built-in reader on each phone and keeps whichever is faster without losing words, so the download is optional.
Folder: ppocr-v6-tiny-finder-v5-readers/ (36.5 MB, 11 files plus files.json).
| File | Size | What it is |
|---|---|---|
manifest.json |
2.4 KB | pack id and version, which file does what, input sizes |
v6-tiny-det.onnx |
1.8 MB | line finder (CPU) |
v6-tiny-det-x1280.tflite |
2.0 MB | line finder for the phone's GPU, five fixed sizes in one file |
v6-tiny-rec.onnx + v6-tiny-rec.txt |
4.5 MB + 27 KB | reader for all text, and its letter list |
v6-tiny-rec-48.tflite |
4.5 MB | the same reader for the GPU, three fixed widths in one file |
v5-en-rec.onnx + v5-en-rec.txt |
7.8 MB + 1.4 KB | English reader used on lines with numbers, and its letter list |
v5-en-rec-48.tflite |
8.0 MB | the same reader for the GPU |
v5-devanagari-rec.onnx + v5-devanagari-rec.txt |
7.9 MB + 1.9 KB | Hindi reader, and its letter list |
files.json |
every file's name, size and SHA-256, for checking a download |
Download
Every file in the folder is needed together. Fetch files.json first, then each file it lists, check
size and SHA-256, and write manifest.json last. With the Hugging Face CLI:
hf download lastabhi/amar-line-readers --include "ppocr-v6-tiny-finder-v5-readers/*" --local-dir amar-line-readers
Plain HTTPS works too: https://huggingface.co/lastabhi/amar-line-readers/resolve/main/ppocr-v6-tiny-finder-v5-readers/files.json
(HTTP Range requests are supported, so a download can resume).
Licence and sources
Apache License 2.0, the licence of the upstream models:
v6-tiny-det: PaddlePaddle/PP-OCRv6_tiny_det_onnxv6-tiny-rec: PaddlePaddle/PP-OCRv6_tiny_rec_onnxv5-en-rec: PaddlePaddle/en_PP-OCRv5_mobile_rec_onnxv5-devanagari-rec: PaddlePaddle/devanagari_PP-OCRv5_mobile_rec_onnx
The ONNX files are the upstream ones. The .tflite files were converted from them by the Amar Vault
project (fixed input sizes, operations rewritten so a phone's GPU can run them, sizes merged into one
file per model). They come with no warranty.
- Downloads last month
- 29
Model tree for lastabhi/amar-line-readers
Base model
PaddlePaddle/PP-OCRv6_tiny_det_onnx