PP-OCR models converted to MNN

All PP-OCR v4/v5/v6 detection/recognition models (mobile/server/tiny/small/medium), language-specific rec models, doc-rec, seal-det, and the PP-LCNet textline-orientation classifier, converted from the official PaddleOCR weights (Apache-2.0) via MNNConvert.

30 .mnn files + registry.json (name → sha256, size, config) consumed by pp-ocr-mnn's runtime auto-downloader.

Do not re-download from PaddleOCR: these are the exact conversions validated by the 811-cell accuracy matrix (CER ≤ 0.05 vs paddle.inference CPU baseline).

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support