LibrePPOCR family notice

This family is a PyTorch port of PP-OCRv5 from PaddleOCR.

Upstream project: https://github.com/PaddlePaddle/PaddleOCR
Upstream license: Apache License 2.0
Reference paper: PaddleOCR 3.0 Technical Report, arXiv:2507.05595

Ported components (from the upstream paths, PaddlePaddle -> PyTorch):
- ppocr/modeling/backbones/rec_lcnetv3.py        -> libreyolo/models/ppocr/backbones.py (PPLCNetV3)
- ppocr/modeling/backbones/rec_pphgnetv2.py      -> libreyolo/models/ppocr/backbones.py (PPHGNetV2B4)
- ppocr/modeling/necks/db_fpn.py                 -> libreyolo/models/ppocr/det.py (RSEFPN, LKPAN)
- ppocr/modeling/necks/intracl.py                -> libreyolo/models/ppocr/det.py (IntraCLBlock)
- ppocr/modeling/heads/det_db_head.py            -> libreyolo/models/ppocr/det.py (DBHead, PFHeadLocal)
- ppocr/modeling/backbones/det_mobilenet_v3.py   -> libreyolo/models/ppocr/det.py (SEModule)
- ppocr/modeling/necks/rnn.py                    -> libreyolo/models/ppocr/rec.py (SequenceEncoder, EncoderWithSVTR)
- ppocr/modeling/backbones/rec_svtrnet.py        -> libreyolo/models/ppocr/rec.py (Attention, Block)
- ppocr/modeling/heads/rec_ctc_head.py           -> libreyolo/models/ppocr/rec.py (CTCHead)
- ppocr/modeling/heads/rec_multi_head.py         -> libreyolo/models/ppocr/rec.py (MultiHead, CTC branch)
- ppocr/postprocess/db_postprocess.py            -> libreyolo/postprocess/ppocr.py (db_postprocess)
- ppocr/postprocess/rec_postprocess.py           -> libreyolo/postprocess/ppocr.py (ctc_decode)
- tools/infer/predict_det.py, predict_rec.py,
  predict_system.py, tools/infer/utility.py      -> libreyolo/models/ppocr/preprocess.py and
                                                    libreyolo/models/ppocr/inference.py
- ppocr/utils/dict/ppocrv5_dict.txt              -> embedded as checkpoint "charset" metadata by
                                                    weights/convert_ppocr_weights.py

Published LibrePPOCR checkpoints are converted from the official PP-OCRv5
training checkpoints (Apache-2.0) released by the PaddleOCR project:
PP-OCRv5_mobile_det, PP-OCRv5_server_det, PP-OCRv5_mobile_rec,
PP-OCRv5_server_rec. Exact source files and their SHA-256 hashes are recorded
on each Hugging Face model card and in the conversion script.

The upstream code and weights are used under the terms of the Apache License,
Version 2.0: http://www.apache.org/licenses/LICENSE-2.0
