PaddleOCR is an open-source library for optical character recognition (OCR) and document parsing tasks. Its latest version, PaddleOCR 3.5, introduces a transformers backend, allowing users to leverage the power of transformer-based models for these tasks. This update enables faster and more accurate processing of documents and images. According to the developers, this change is expected to improve performance in tasks such as text recognition and document layout analysis. --- Why it matters: This matters to engineers working on AI applications that involve document parsing or OCR, as it provides a new tool for improving their models' accuracy and efficiency. Source: https://huggingface.co/blog/PaddlePaddle/paddleocr-transformers