Views
No views yet
| Engine | CER |
|---|---|
| V8+LM (this model + KenLM post-processing) | 0.20% |
| V8 (this model, raw) | 0.26% |
| V7 (previous) | 1.02% |
| Tesseract (mkd) | 1.43% |
| Qwen 3.5 122B (best VLM) | 0.39% |
1from paddleocr import PaddleOCR
2ocr = PaddleOCR(
3 text_detection_model_name='PP-OCRv6_medium_det',
4 text_recognition_model_name='PP-OCRv5_server_rec',
5 text_recognition_model_dir='./mk_rec_v8_infer',
6 text_det_thresh=0.15,
7 text_det_box_thresh=0.28,
8 text_det_unclip_ratio=3.0,
9 use_doc_orientation_classify=False,
10 use_doc_unwarping=False,
11 use_textline_orientation=False,
12)
13result = ocr.predict('your_page.jpg')
14for page in result:
15 for text, score in zip(page['rec_texts'], page['rec_scores']):
16 print(f'{score:.2f} {text}')1pip install huggingface_hub
2python3 -c "
3from huggingface_hub import snapshot_download
4snapshot_download('mjurukov/macedonian-ocr-v8', local_dir='./mk_rec_v8_infer')
5"