PaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers Backend

PaddleOCR 3.5 introduces a Transformers-based backend for optical character recognition (OCR) and document parsing tasks, enabling improved performance and integration with the Hugging Face ecosystem. This release leverages transformer models to enhance text detection, recognition, and document understanding capabilities compared to previous CNN-based approaches. The update demonstrates PaddlePaddle's effort to modernize its OCR framework with state-of-the-art deep learning architectures available through the Transformers library.

Source: Hugging Face | 2026-05-18