PP-OCRv6 on Hugging Face: 50-Language OCR from 1.5M to 34.5M Parameters

PP-OCRv6 is the latest generation of PaddleOCR’s universal optical character recognition model family, designed for real-world text detection and recognition across documents, screenshots, and multilingual images. The model family scales from 1.5M to 34.5M parameters across three tiers (tiny, small, and medium) to balance performance with deployment constraints. The medium and small tiers support 50 languages, with the medium version achieving significant accuracy improvements over its predecessor. Key architectural innovations include a unified PPLCNetV4 backbone and a RepLKFPN module for text detection.