pptspine
Copyright 2026 pptspine authors

Licensed under the Apache License, Version 2.0.
You may obtain a copy of the License in the accompanying LICENSE file or at
http://www.apache.org/licenses/LICENSE-2.0.

------------------------------------------------------------------------------
THIRD-PARTY BUNDLED DATA ATTRIBUTIONS
------------------------------------------------------------------------------

The published `pptspine` wheel bundles the PaddleOCR PP-OCRv5 ONNX model weights
(inside the package at `pptspine/_models`) so that a bare `pip install pptspine`
is full-OCR-capable offline, with no separate data package and no network. The
OCR inference runs in the sibling family crate `ocrspine`; the authoritative,
per-file provenance and conversion record are retained in
`pptspine/_models/PROVENANCE.md`.

* PaddleOCR PP-OCR models and recognition dictionary
  (`pptspine/_models/ppocrv5_det.onnx`,
   `pptspine/_models/ppocrv5_rec.onnx`,
   `pptspine/_models/ppocrv5_cls.onnx`,
   `pptspine/_models/ppocr_keys_v5.txt`)
  Copyright (c) PaddlePaddle Authors.
  Licensed under the Apache License, Version 2.0.
  The bundled files are the PP-OCRv5 mobile text-detection and text-recognition
  models, the PP-LCNet text-line-orientation classification model, and the
  PP-OCRv5 recognition character dictionary, converted to ONNX (via Paddle2ONNX)
  from the upstream PaddlePaddle weights, then post-processed by a deterministic
  ONNX dim-name/shape-hint cleanup so the pure-Rust `tract` runtime can parse
  them (no weight changes).
  Source: https://github.com/PaddlePaddle/PaddleOCR
  Per-file provenance and the conversion record are retained in
  `pptspine/_models/PROVENANCE.md`.

Third-party Rust dependencies are distributed under their own permissive
licenses (MIT, Apache-2.0, BSD, Zlib, ISC, Unicode, and similar). Apache-2.0 is
compatible with each.
