Latin PP-OCRv5 Mobile Rec
PaddlePaddle/latin_PP-OCRv5_mobile_rec
published Jul 2025 · updated Oct 2025
Latin PP-OCRv5 Mobile Rec is an image-to-text model that recognizes Latin alphabet text lines from images, achieving 84.7% accuracy.
specs
| Task | Image-to-Text |
| Architecture | PP-OCRv5 Mobile |
| Accuracy | 84.7% on Latin language dataset |
| License | Apache 2.0 |
about this model
latin_PP-OCRv5_mobile_rec is an image-to-text model for text line recognition, part of the PP-OCRv5 series developed by the PaddleOCR team. It is designed to efficiently and accurately recognize Korean text from images. The model achieves an accuracy of 84.7% on a Latin alphabet language dataset, as reported by the PaddleOCR team. Notably, the evaluation uses a strict character-level metric: if any character (including punctuation) in a line is incorrect, the entire line is marked as wrong, ensuring higher practical accuracy.
| Model | Latin Alphabet Language Dataset Accuracy (%) |
|---|---|
| latin_PP-OCRv5_mobile_rec | 84.7 |
Sample Recognition Output
The following image shows the recognition result for a sample input:
Pipeline Visualization
When used within the PP-OCRv5 pipeline (including text detection and orientation classification), the model produces end-to-end OCR results as illustrated below:
best for
- ·Digitizing printed Latin documents
- ·OCR for scanned book pages
- ·Real-time text extraction from images
FAQ
It is best for recognizing Latin script text lines from images, particularly in document digitization and printed text OCR.
It is released under the Apache 2.0 license.
The model expects images containing text lines; it can be used via PaddleOCR or the gigarouter API.
Use the gigarouter OpenAI-compatible endpoint with your API key, sending an image URL or base64-encoded image.
The model achieves 84.7% accuracy on a Latin language dataset, with line-level evaluation.
We're benchmarking and onboarding Latin PP-OCRv5 Mobile Rec as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.