PaddleOCR

PolyOCR Service is a community-driven independent service designed to provide convenient API deployment for PaddleOCR 3.x. Its core capabilities include: - **Multi-language OCR**: Supports 78 languages (covering Chinese, Japanese, Korean, Latin, Cyrillic, Arabic, etc.), with support for specifying languages via language codes, English names, or Chinese names. - **Translation Integration**: Provides optional translation interfaces, supporting OpenAI-compatible translation services, with limits on the number of input entries and characters. - **PaddleOCR-VL**: Provides a standalone vision-language model service, requiring an independent API Key and accepting only file uploads. - **Robustness and Validation**: Performs language validation at request boundaries, with pre-validation for image byte size, pixel count, and confidence thresholds; synchronous inference is executed via a thread pool and concurrency is limited by semaphores. - **Benchmarking**: The project includes built-in detailed benchmarking tools for accuracy, robustness (targeting degradation scenarios such as blur, compression, and rotation), and real-photo tests (based on the CORD-v2 dataset). The service provides a Web page and REST API, supports Docker deployment, and is licensed under the Apache License 2.0.