Skip to content

Releases: PaddleOCRCore/CoreOCROnnx

v4.2.1

Choose a tag to compare

@PaddleOCRCore PaddleOCRCore released this 22 Sep 02:55

CoreOCROnnx v4.2.1

优化

  • 升级 PaddleOCROnnx 4.2.1 运行库,进一步提升 OCR 识别性能,识别速度大幅提升。
  • 优化 CPU 默认线程配置,默认8为最佳实践,提升常用场景下的识别效率。
  • 支持配置 OCR 引擎实例数量,便于提高并发识别吞吐量。
  • 支持模型及图片中文路径

新增

  • 新增单字坐标返回能力,可按需输出字符级识别框信息。
  • 更新 C++、C#、Go、Python 示例及 WinForms Demo,适配最新运行库与模型配置。

修复

  • 改进 WinForms 批量识别的异步处理与日志输出表现。

v4.1.0

Choose a tag to compare

@PaddleOCRCore PaddleOCRCore released this 16 Jul 02:21

feat:增加后端TensorRT支持。
fix:FreeResultBuffer增加自动回收,智能缓冲区管理
发布Nuget Runtime 包:

NuGet 包 平台/架构 推理设备 主要模型格式 License 要求
CoreOCRRuntime.Onnx.CPU.win-x86 Windows x86 CPU .onnx 不需要
CoreOCRRuntime.Onnx.CPU.win-x64 Windows x64 CPU .onnx 不需要
CoreOCRRuntime.Onnx.CPU.linux-x64 Linux x64 CPU .onnx 需要
CoreOCRRuntime.DirectML.win-x86 Windows x86 DirectML GPU,也可使用 CPU .onnx CPU 不需要;GPU 需要
CoreOCRRuntime.DirectML.win-x64 Windows x64 DirectML GPU,也可使用 CPU .onnx CPU 不需要;GPU 需要
CoreOCRRuntime.OpenVino.CPU.win-x64 Windows x64 CPU .onnx、.xml + .bin 不需要
CoreOCRRuntime.OpenVino.GPU.win-x64 Windows x64 Intel GPU,也包含 CPU 插件 .onnx、.xml + .bin CPU 不需要;GPU 需要
CoreOCRRuntime.TensorRT.GPU.win-x64 Windows x64 NVIDIA GPU .onnx、.engine、.plan 需要 GPU 授权
CoreOCRRuntime.TensorRT.GPU.linux-x64 Linux x64 NVIDIA GPU .onnx、.engine、.plan 需要 GPU 授权

v4.0.0

Choose a tag to compare

@PaddleOCRCore PaddleOCRCore released this 07 Jun 12:00

OCRRuntimeOnnx:基于onnxruntime,支持PaddleOCR及Yolo,支持onnx格式模型
OCRRuntimeOpenVino:基于OpenVino,支持PaddleOCR及Yolo,同时支持onnx及IR格式模型

ONNX:.onnx,可直接 read_model/compile_model
OpenVINO IR:.xml + .bin,官方推荐的生产部署格式