Skip to content

v1.0.0 lw.PPOCR.Inference — ABI冻结 · 四种后端 · 任意语言

Choose a tag to compare

View all tags
@lxw112190 lxw112190 released this 16 Jul 01:40
v1.0.0
5504aad

lw.PPOCR.Inference v1.0.0 permanently freezes the public C ABI, the private
Runtime ABI, and the model manifest schema v1.

Stability promises

  • Public C ABI v1 — frozen. No function removed, no signature changed.
    Structs and enums grow append-only. LW_PPOCR_API_VERSION = 1 (permanent).
  • Private Runtime ABI v1 — frozen.
  • Model manifest schema v1 — frozen.
  • Threading & memory contract — frozen.
  • Security fixes backported for ≥ 2 years.
  • See docs/abi-v1-stability.md for the full guarantee.

Four backends, one API

Backend Device
OpenCV DNN CPU
ONNX Runtime DirectML CPU / GPU (Intel, AMD, NVIDIA)
OpenVINO CPU
TensorRT FP16 NVIDIA GPU

What's in v1.0.0

  • C ABI permanently frozen with long-term stability commitment
  • Model manifest schema v1 frozen with $id and full documentation
  • Golden correctness test suite (text, bounding boxes, scores)
  • C example (LoadLibrary + GetProcAddress, zero dependencies)
  • Python binding example (ctypes, with statement, run_file())
  • GitHub Actions CI (VS build + ABI smoke/stability + golden test)
  • Backend split packaging (-Split / -Runtime)
  • Compatibility matrix across OS, GPU, driver, and SDK versions
  • Third-party license audit (clipper BSL-1.0, OpenCV/OpenVINO Apache-2.0,
    ONNX Runtime/DirectML MIT, TensorRT/CUDA NVIDIA EULA)

Any language

C · C++ · C# · Python · Delphi · Go · Rust · Java · ...
    → lw.PPOCR.dll          (stable C ABI)
        → OpenCV DNN / DirectML / OpenVINO / TensorRT
Assets 4
Loading