Bina 0.1.0: README.md
Browse files
README.md
CHANGED
|
@@ -1,58 +1,58 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: other
|
| 3 |
-
license_name: bina-desktop-and-ai-pubs-openrail-m
|
| 4 |
-
language:
|
| 5 |
-
- fa
|
| 6 |
-
tags:
|
| 7 |
-
- ocr
|
| 8 |
-
- persian
|
| 9 |
-
- farsi
|
| 10 |
-
- windows
|
| 11 |
-
- offline
|
| 12 |
-
- desktop-app
|
| 13 |
-
pipeline_tag: image-text-to-text
|
| 14 |
-
base_model: Reza2kn/Bina-0.3-Koochik
|
| 15 |
-
---
|
| 16 |
-
|
| 17 |
-
# Bina (بینا) 0.1 — offline Persian OCR for Windows
|
| 18 |
-
|
| 19 |
-
**Bina** turns photos, scans and PDFs of Persian text into editable text, right on your own computer. There is no internet connection, no account, and no upload. Your files never leave the machine.
|
| 20 |
-
|
| 21 |
-
Drop in a page and Bina reads it, then shows the text **next to the original page** so you can compare, correct and copy it.
|
| 22 |
-
|
| 23 |
-
## Download
|
| 24 |
-
| File | Size | For |
|
| 25 |
-
|---|---|---|
|
| 26 |
-
| `Bina-Setup-0.1.0.exe` | ~1.2 GB | Windows 10 (1809+) / 11, 64-bit |
|
| 27 |
-
|
| 28 |
-
SHA-256 checksums are in `SHA256SUMS.txt`.
|
| 29 |
-
|
| 30 |
-
**Install:** run the file. No administrator rights are needed (it installs for your user only). The installer is not code-signed yet, so Windows SmartScreen may show "Windows protected your PC". Click **More info → Run anyway**.
|
| 31 |
-
|
| 32 |
-
## Features
|
| 33 |
-
- Images (JPG, PNG, WEBP, TIFF, BMP) and multi-page PDFs, plus pasting a screenshot with Ctrl+V
|
| 34 |
-
- Text and page side by side, with zoom and a full-screen view, to check every line
|
| 35 |
-
- Live text as it is being read; stop at any time and read again
|
| 36 |
-
- Copy, or save as UTF-8 `.txt` (one file, or all files together)
|
| 37 |
-
- Three reading qualities (fast / normal / precise) and a CPU-use setting
|
| 38 |
-
- Persian right-to-left interface; light, dark or follows Windows
|
| 39 |
-
- Runs on the CPU, so no graphics card is needed. Uses about 3 GB of memory while open, and all of it is returned the moment you close the window.
|
| 40 |
-
|
| 41 |
-
## System requirements
|
| 42 |
-
- Windows 10 1809 or newer, 64-bit
|
| 43 |
-
- 8 GB of memory recommended (the model uses about 3 GB)
|
| 44 |
-
- About 3 GB of free disk space
|
| 45 |
-
- Speed depends on the processor. A dense page takes from about one to a few minutes on a typical laptop.
|
| 46 |
-
|
| 47 |
-
## Model
|
| 48 |
-
Bina uses **[Bina-0.3-Koochik](https://huggingface.co/Reza2kn/Bina-0.3-Koochik)**, a Persian full-page OCR model developed by **[Reza Sayar](https://huggingface.co/Reza2kn)**. It is distributed under the AI Pubs OpenRAIL-M licence (modified), and its use restrictions apply to everyone who uses this program. The full licence is installed with the program (`model\LICENSE`).
|
| 49 |
-
|
| 50 |
-
Known limits: complex layouts (multi-column pages, poetry in two columns, tables) and low-quality photos can come out incomplete or out of order. Bina warns you when that happens.
|
| 51 |
-
|
| 52 |
-
## Credits
|
| 53 |
-
- Application: designed and developed by **[
|
| 54 |
-
- OCR model: **Reza Sayar** ([Bina-0.3-Koochik](https://huggingface.co/Reza2kn/Bina-0.3-Koochik))
|
| 55 |
-
- Built with Python, PyTorch, Hugging Face Transformers, Electron, PDFium and the Vazirmatn font (each under its own open licence, listed in `NOTICE.txt`)
|
| 56 |
-
|
| 57 |
-
## Privacy
|
| 58 |
-
Bina makes no network connections. The engine listens only on 127.0.0.1 (this computer), and the window blocks every request that is not that engine. No telemetry, no update checks, no logs leave the computer.
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: bina-desktop-and-ai-pubs-openrail-m
|
| 4 |
+
language:
|
| 5 |
+
- fa
|
| 6 |
+
tags:
|
| 7 |
+
- ocr
|
| 8 |
+
- persian
|
| 9 |
+
- farsi
|
| 10 |
+
- windows
|
| 11 |
+
- offline
|
| 12 |
+
- desktop-app
|
| 13 |
+
pipeline_tag: image-text-to-text
|
| 14 |
+
base_model: Reza2kn/Bina-0.3-Koochik
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# Bina (بینا) 0.1 — offline Persian OCR for Windows
|
| 18 |
+
|
| 19 |
+
**Bina** turns photos, scans and PDFs of Persian text into editable text, right on your own computer. There is no internet connection, no account, and no upload. Your files never leave the machine.
|
| 20 |
+
|
| 21 |
+
Drop in a page and Bina reads it, then shows the text **next to the original page** so you can compare, correct and copy it.
|
| 22 |
+
|
| 23 |
+
## Download
|
| 24 |
+
| File | Size | For |
|
| 25 |
+
|---|---|---|
|
| 26 |
+
| `Bina-Setup-0.1.0.exe` | ~1.2 GB | Windows 10 (1809+) / 11, 64-bit |
|
| 27 |
+
|
| 28 |
+
SHA-256 checksums are in `SHA256SUMS.txt`.
|
| 29 |
+
|
| 30 |
+
**Install:** run the file. No administrator rights are needed (it installs for your user only). The installer is not code-signed yet, so Windows SmartScreen may show "Windows protected your PC". Click **More info → Run anyway**.
|
| 31 |
+
|
| 32 |
+
## Features
|
| 33 |
+
- Images (JPG, PNG, WEBP, TIFF, BMP) and multi-page PDFs, plus pasting a screenshot with Ctrl+V
|
| 34 |
+
- Text and page side by side, with zoom and a full-screen view, to check every line
|
| 35 |
+
- Live text as it is being read; stop at any time and read again
|
| 36 |
+
- Copy, or save as UTF-8 `.txt` (one file, or all files together)
|
| 37 |
+
- Three reading qualities (fast / normal / precise) and a CPU-use setting
|
| 38 |
+
- Persian right-to-left interface; light, dark or follows Windows
|
| 39 |
+
- Runs on the CPU, so no graphics card is needed. Uses about 3 GB of memory while open, and all of it is returned the moment you close the window.
|
| 40 |
+
|
| 41 |
+
## System requirements
|
| 42 |
+
- Windows 10 1809 or newer, 64-bit
|
| 43 |
+
- 8 GB of memory recommended (the model uses about 3 GB)
|
| 44 |
+
- About 3 GB of free disk space
|
| 45 |
+
- Speed depends on the processor. A dense page takes from about one to a few minutes on a typical laptop.
|
| 46 |
+
|
| 47 |
+
## Model
|
| 48 |
+
Bina uses **[Bina-0.3-Koochik](https://huggingface.co/Reza2kn/Bina-0.3-Koochik)**, a Persian full-page OCR model developed by **[Reza Sayar](https://huggingface.co/Reza2kn)**. It is distributed under the AI Pubs OpenRAIL-M licence (modified), and its use restrictions apply to everyone who uses this program. The full licence is installed with the program (`model\LICENSE`).
|
| 49 |
+
|
| 50 |
+
Known limits: complex layouts (multi-column pages, poetry in two columns, tables) and low-quality photos can come out incomplete or out of order. Bina warns you when that happens.
|
| 51 |
+
|
| 52 |
+
## Credits
|
| 53 |
+
- Application: designed and developed by **[Khosrow Reza](https://huggingface.co/rezkhos)**
|
| 54 |
+
- OCR model: **Reza Sayar** ([Bina-0.3-Koochik](https://huggingface.co/Reza2kn/Bina-0.3-Koochik))
|
| 55 |
+
- Built with Python, PyTorch, Hugging Face Transformers, Electron, PDFium and the Vazirmatn font (each under its own open licence, listed in `NOTICE.txt`)
|
| 56 |
+
|
| 57 |
+
## Privacy
|
| 58 |
+
Bina makes no network connections. The engine listens only on 127.0.0.1 (this computer), and the window blocks every request that is not that engine. No telemetry, no update checks, no logs leave the computer.
|