skip to content
gigarouter gigarouter
models / image-to-text · coming soon

PP-OCRv5 Server Rec

PaddlePaddle/PP-OCRv5_server_rec

published Jun 2025 · updated Jul 2025

PP-OCRv5 Server Rec is an image-to-text model that recognizes text from images, supporting Simplified Chinese, Traditional Chinese, English, Japanese, and complex scenarios such as handwriting, vertical text, pinyin, and rare characters using a single model.

status
coming soon
API providers
0
downloads / mo
189.4K
license
apache-2.0

specs

TaskImage-to-Text (Text Recognition)
ArchitecturePP-OCRv5 server recognition model
Average Accuracy0.8401 (line-level over 12 scenarios)

about this model

PP-OCRv5_server_rec is an image-to-text recognition model that efficiently supports simplified Chinese, traditional Chinese, English, and Japanese, along with handwriting, vertical text, pinyin, and rare characters using a single model. It achieves strong accuracy across twelve diverse scenarios, evaluated with a strict line-level metric where any character or punctuation error marks the entire line as incorrect. The overall average accuracy is 0.8401.
ScenarioAccuracy
Handwritten Chinese0.5807
Handwritten English0.5806
Printed Chinese0.9013
Printed English0.8679
Traditional Chinese0.7472
Ancient Text0.6039
Japanese0.7372
General Scenario0.5946
Pinyin0.8384
Rotation0.7435
Distortion0.9314
Artistic Text0.6397
The visualized recognition result on a sample image shows the model outputting the text “day as a reminder of the” with a confidence score of 0.9535. Sample image with recognized text 'day as a reminder of the' overlayed This model is hosted on gigarouter as a managed, OpenAI-compatible API, providing reliable image-to-text inference without local installation.

best for

FAQ

What languages does PP-OCRv5 Server Rec support?

It supports Simplified Chinese, Traditional Chinese, English, Japanese, as well as handwriting, vertical text, pinyin, and rare characters.

What is the average accuracy of this model?

It achieves an average line-level accuracy of 0.8401 across 12 scenarios, including printed, handwritten, and distorted text.

How can I call this model via the gigarouter API?

Use the gigarouter OpenAI-compatible endpoint with an API key. Input an image and receive recognized text as the response.

What input format does the model accept?

The model accepts image URLs or local image files. In the pipeline, it processes cropped text-line images.

Is this model part of a larger OCR pipeline?

Yes, it can be used standalone for text recognition or as part of the PP-OCRv5 pipeline that includes detection and orientation modules.

not yet live

We're benchmarking and onboarding PP-OCRv5 Server 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.

related image-to-text models

compare all →