PP-OCRv5_mobile_det
PaddlePaddle/PP-OCRv5_mobile_det
A popular open image-to-text model, with 129.4K downloads a month. gigarouter benchmarks and hosts it as an OpenAI-compatible API.
about this model
PP-OCRv5_mobile_det is a text detection model that identifies and localizes text regions in images, supporting handwritten, printed, vertical, rotated, curved, and artistic text across multiple languages including Simplified Chinese, Traditional Chinese, English, and Japanese.
Key Strengths
The model is designed for robust handling of complex layouts, varying text sizes, and challenging backgrounds. It is suitable for practical applications such as document analysis, license plate recognition, and scene text detection.
Performance Benchmarks
The following table shows per-category accuracy (F1-score) on the benchmark dataset:
| Scenario | Accuracy |
|---|---|
| Handwritten Chinese | 0.744 |
| Handwritten English | 0.777 |
| Printed Chinese | 0.905 |
| Printed English | 0.910 |
| Traditional Chinese | 0.823 |
| Ancient Text | 0.581 |
| Japanese | 0.727 |
| General Scenario | 0.721 |
| Pinyin | 0.575 |
| Rotation | 0.647 |
| Distortion | 0.827 |
| Artistic Text | 0.525 |
| Average | 0.770 |
The model excels on printed text (Chinese and English above 0.90) and distorted text (0.827), while maintaining useful performance on handwriting and rotation scenarios.
Integration via Gigarouter
As a hosted API, developers can call PP-OCRv5_mobile_det directly using an OpenAI-compatible endpoint, with no local installation required.
Example Detection Output
Detected polygons and confidence scores are returned as structured JSON. The following image shows a typical detection result:
An example pipeline output combining detection with recognition is shown below:
We're benchmarking and onboarding PP-OCRv5_mobile_det 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.