Baidu PaddleOCR open-source sixth-generation OCR model PP-OCRv6

Baidu PaddlePaddle officially launched PP-OCRv6, introducing three model levels—Tiny, Small, and Medium—for the first time. The model's detection and recognition accuracy are improved by 4.9% and 5.1% respectively compared to v5, with CPU inference speed up to 5.2 times faster than its predecessor. The model supports 50 languages and adds industrial scenarios such as circuit boards and digital tube displays, boasting the highest robustness ever achieved. The model is now available on the official website and open-sourced on GitHub and HuggingFace.