Open insinfo opened 3 months ago
perhaps training on this dataset could help improve accuracy https://zenodo.org/records/7872951
I think PP-OCRv4 is still good for clear text on the image, the text below the example image is a bit blurry, which may be the reason for the poor recognition.
This issue is stale because it has been open for 90 days with no activity.
🔎 Search before asking
🐛 Bug (问题描述)
I did a test to OCR scanned documents in Brazilian Portuguese, and I saw that PaddleOCR makes a lot of mistakes on scanned documents in Portuguese, I used the C# implementation with the larger server models ch_PP-OCRv4_det_server_infer, ch_ppocr_mobile_v2.0_cls_infer, ch_PP-OCRv4_rec_server_infer
https://github.com/raoyutian/PaddleOCRSharp
example 1
result:
the correct thing would be
example 2
result:
the correct thing would be
example 3
result:
the correct thing would be
🏃♂️ Environment (运行环境)
Windows 11 .NET C#
🌰 Minimal Reproducible Example (最小可复现问题的Demo)