| # Bina 0.2 Rizeh: quick start / راهاندازی سریع |
|
|
| This repository includes both the Persian recognizer and the PP-OCRv6 Medium detector, so the commands below do not download model weights at runtime. |
|
|
| این مخزن هم مدل تشخیص متن فارسی و هم آشکارساز PP-OCRv6 Medium را دارد؛ بنابراین هنگام اجرا وزن دیگری دانلود نمیشود. |
|
|
| ## Install / نصب |
|
|
| ```bash |
| git lfs install |
| git clone https://huggingface.co/Reza2kn/Bina-0.2-Rizeh |
| cd Bina-0.2-Rizeh |
| python -m pip install "paddlepaddle>=3.2,<4" "paddleocr>=3.3,<4" |
| ``` |
|
|
| The command above installs CPU PaddlePaddle. For NVIDIA GPUs, install the PaddlePaddle GPU wheel matching your CUDA version first, then install `paddleocr>=3.3,<4`. |
|
|
| دستور بالا نسخه CPU را نصب میکند. برای کارت NVIDIA ابتدا نسخه GPU سازگار با CUDA خود را نصب کنید و سپس PaddleOCR را نصب کنید. |
|
|
| ## Full page / صفحه کامل |
|
|
| ```bash |
| python bina_page_ocr.py page.jpg --device cpu --output result.json |
| # NVIDIA example: |
| python bina_page_ocr.py page.jpg --device gpu:0 --output result.json |
| ``` |
|
|
| Images and PDFs are accepted. Output is UTF-8 JSON containing page text, logical-order Persian lines, confidence scores, and boxes. |
|
|
| ## Line crops / برش خط |
|
|
| ```bash |
| python bina_text_recognition.py line.jpg --device cpu |
| ``` |
|
|
| Python: |
|
|
| ```python |
| from bina_text_recognition import BinaTextRecognition |
| model = BinaTextRecognition(device="cpu") |
| for row in model.predict("line.jpg"): |
| print(row["text"], row["score"]) |
| ``` |
|
|
| ## Paths / مسیرها |
|
|
| - Recognizer: `inference/` |
| - Bundled detector: `detector/` |
| - Best training checkpoint: `checkpoint/` |
| - Exact benchmark evidence: `benchmark/` |
|
|