PaddleOCR As a text recognition development kit under the PaddlePaddle deep learning open source framework, v2.8.0 has released a milestone update. This version introduces cutting-edge OCR technology, including the champion solution of the PaddleOCR algorithm model challenge, such as the scene text recognition algorithm SVTRv2 andTable recognition algorithmSLANet-LCNetV2 sets a new standard in the OCR field.
At the same time, the project structure has been deeply optimized, and non-core modules have been moved to the new warehouse, allowing the project to focus more on OCR core technology. In addition, historical problems such as the model failing to run after updating Backbone, numpy version dependency conflicts, and Mac system running freezes have been solved, improving the user experience.
The new version also includes a fix for the loss of OCR results in layout analysis, the introduction of pyproject.toml to comply with the PEP518 specification, and optimization improvements such as sliding window operations for large-scale image reasoning, which enhances the stability, compatibility, and performance of the software. The support and contribution of the open source community are crucial to every progress of PaddleOCR v2.8.0, and the efforts of PMC members and contributors are particularly appreciated.
PaddleOCR is currently building a dedicated site for document tutorials, which will provide keyword search capabilities and an elegant and comfortable interface.
Project address:https://github.com/PaddlePaddle/PaddleOCR