PaddleOCR Releases New Model PP-OCRv6

PaddleOCR's new OCR model series, PP-OCRv6, has been officially released. This update consists of Tiny, Small, Medium models ranging from 1.5M to 34.5M parameters, offering optimized deployment options for various environments including browsers, edge devices, and servers.

Key features are as follows:

• Performance Improvement: Compared to the previous version, PP-OCRv5, Detection accuracy improved by 4.9% and Recognition accuracy improved by 5.1%.
• Inference Speed: CPU inference speed is up to 5.2x faster when using OpenVINO.
• Versatility: A single unified model supports 50 languages, and it also supports special scenarios such as PCBs, CAD drawings, digital tubes, and dot-matrix text.
• License: Released under the Apache 2.0 license, making commercial use easy.