mirror of
https://github.com/RapidAI/RapidOCR.git
synced 2026-11-10 21:10:55 +08:00
Cursor AgentandDaniel
3622c4598e
Expose the PP-OCRv6 engine through a stable C ABI and installable shared and static libraries, plus a ctypes Python package that follows the RapidOCR call shape. Keep AVX-512 out of the AVX2 object file and speed up GEMM accumulate and the AArch64 GEMM tile. Co-authored-by: Daniel <znsoft@163.com>
rapidocr (native PP-OCRv6)
Python package for the V5_NG C++ engine. The import surface follows
rapidocr on main:
from rapidocr import RapidOCR
engine = RapidOCR(params={
"Det.model_path": "PP-OCRv6_det_tiny.onnx",
"Rec.model_path": "PP-OCRv6_rec_tiny.onnx",
"Rec.rec_keys_path": "ppocrv6_tiny_dict.txt",
"Global.use_cls": False,
})
result = engine("page.png")
print(result.txts, result.scores)
print(result.to_json())
If model paths are omitted, the package downloads the PP-OCRv6 ONNX bundle
for Det.model_type / Rec.model_type (tiny, small, or medium).
Build a wheel from a checkout that already has the Vulkan-Headers submodule:
pip install scikit-build-core
pip install ./python
The binding is ctypes over the C ABI (include/ppocr/ppocr.h), so the wheel
ships libppocr and does not add a second C++ wrapper.
See docs/API.md for the differences from main (no angle classifier, no
word boxes, PP-OCRv6 only).