Full PP-OCRv5 PyTorch model zoo converted from official PaddlePaddle .pdparams:
- 2 text detection models (mobile/server)
- 2 base text recognition models (zh/zh_tra/en/ja, mobile/server)
- 11 multilingual rec models (en/korean/latin/eslav/cyrillic/arabic/devanagari/th/el/ta/te), 100+ languages total
- All safetensors, bit-exact with original PaddleOCR .pdparams outputs
- Includes configs/ and dicts/ needed for inference
- Completely rewritten README (multilingual coverage + usage snippets)
