mirror of
https://github.com/deepinsight/insightface.git
synced 2025-12-30 08:02:27 +00:00
5.4 KiB
5.4 KiB
Overview of CMake Option for Compilation
Here are the translation details for the compilation parameters as per your requirement:
| Parameter | Default Value | Description |
|---|---|---|
| ISF_THIRD_PARTY_DIR | 3rdparty | Path for required third-party libraries |
| ISF_SANITIZE_ADDRESS | OFF | Enable AddressSanitizer for memory error detection |
| ISF_SANITIZE_LEAK | OFF | Enable LeakSanitizer to detect memory leaks |
| ISF_ENABLE_SYMBOL_HIDING | ON | Enable symbol hiding by default for better security and performance. Only symbols explicitly marked for export will be visible. |
| ISF_INSTALL_CPP_HEADER | OFF | Whether to install the headers file for CPP-API (the default C-API with wider compatibility is recommended) |
| ISF_ENABLE_RKNN | OFF | Enable RKNN for Rockchip embedded devices |
| ISF_RK_DEVICE_TYPE | RV1109RV1126 | Target device model for Rockchip(Supports RV1109RV1126, RV1106, RV356X) |
| ISF_RK_COMPILER_TYPE | armhf | The armhf, armhf-uclibc and aarch64 compilers are supported. Select one based on the actual situation |
| ISF_ENABLE_RGA | OFF | Enable RGA image acceleration on Rockchip devices (currently only supported on devices using RKNPU2) |
| ISF_BUILD_LINUX_ARM7 | OFF | Compile for ARM7 architecture |
| ISF_BUILD_LINUX_AARCH64 | OFF | Compile for AARCH64 architecture |
| ISF_BUILD_WITH_TEST | OFF | Compile test case programs |
| ISF_BUILD_WITH_SAMPLE | ON | Compile sample programs |
| ISF_BUILD_SHARED_LIBS | ON | Compile shared libraries |
| ISF_ENABLE_BENCHMARK | OFF | Enable Benchmark tests for test cases |
| ISF_ENABLE_USE_LFW_DATA | OFF | Enable using LFW data for test cases |
| ISF_ENABLE_TEST_EVALUATION | OFF | Enable evaluation functionality for test cases, must be used together with ISF_ENABLE_USE_LFW_DATA |
| ISF_ENABLE_TEST_INTERNAL | OFF | Enable test cases for some internal functions, requires disabling ISF_ENABLE_SYMBOL_HIDING for compilation. |
| ISF_BUILD_SAMPLE_INTERNAL | OFF | Enable executable examples for some internal functions, requires disabling ISF_ENABLE_SYMBOL_HIDING for compilation. |
| ISF_ENABLE_TENSORRT | OFF | Enable the backend of inference using TensorRT, Linux must be on NVIDIA devices, and CUDA, TensorRT-10 must be installed |
| TENSORRT_ROOT | /usr/local/TensorRT | TensorRT-10 installation path |
| ISF_GLOBAL_INFERENCE_BACKEND_USE_MNN_CUDA | OFF | Enable global MNN_CUDA inference mode, requires device support for CUDA |
| ISF_LINUX_MNN_CUDA | "" | Specific MNN library path, requires pre-compiled MNN library supporting MNN_CUDA, only effective when ISF_GLOBAL_INFERENCE_BACKEND_USE_MNN_CUDA is enabled |
| ISF_ENABLE_APPLE_EXTENSION | OFF | If you are using an Apple device (MacOS/iOS), you can enable it to allow some parts of the SDK's models to switch to certain backend neural network acceleration inference on Apple devices, such as Metal and ANE |
| ISF_MNN_CUSTOM_SOURCE | "" | Using this option to replace different versions of MNN, the input must be the root directory of the MNN source code |
| INSPIRECV_BACKEND_OPENCV | OFF | The image processing backend relies on OpenCV and is not recommended |
| ISF_ENABLE_COST_TIME | OFF | This parameter can be used to print the time of some important compute nodes in the Debug state |
| ISF_NEVER_USE_OPENCV | ON | When you need to use OpenCV as the image processing engine, you need to turn this option off |