88Z AI, 15 June — Baidu PaddlePaddle has released PP-OCRv6, offering Tiny, Small, and Medium variants for the first time. Detection and recognition accuracy have improved by 4.9% and 5.1% respectively over the previous version, with CPU inference speed up to 5.2 times faster. The new model supports 50 languages, adds industrial scenarios like circuit boards and digital tubes, and features enhanced robustness. It is now available on the official website and open-sourced on GitHub and HuggingFace.
At a glance
- PP-OCRv6 introduces three model sizes with CPU inference speed up to 5.2x faster
- Supports 50 languages and industrial scenarios with significantly improved accuracy
The update
Baidu PaddlePaddle has officially launched PP-OCRv6, the sixth generation of its optical character recognition model. This update introduces Tiny, Small, and Medium model architectures for the first time, catering to diverse deployment needs. In terms of performance, detection accuracy has improved by 4.9% and recognition accuracy by 5.1% compared to v5, while CPU inference speed is up to 5.2 times faster, significantly optimizing operational efficiency. PP-OCRv6 expands language support to 50 languages and is specifically optimized for complex industrial scenarios such as circuit boards and digital tubes, achieving its highest robustness to date. The model is now live on the official Baidu PaddlePaddle website and has been open-sourced on GitHub and HuggingFace, allowing developers to access and use it directly.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 飞桨 PaddlePaddleSource published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.