| # VisDrone YOLO11s C++ SDK |
|
|
| 预编译的 C++ 推理库和可执行程序,基于 libdet.axera。 |
|
|
| ## 文件说明 |
| ``` |
| bin/visdrone_detect 可执行程序(aarch64,开箱即用) |
| lib/libdet.so 共享库(aarch64,OpenCV 静态链接) |
| include/libdet.h C API 头文件 |
| include/ax_devices.h 设备枚举头文件 |
| ``` |
|
|
| ## 运行依赖 |
| 板端运行时库(板端已预装): |
| - `/soc/lib/libax_engine.so` |
| - `/soc/lib/libax_sys.so` |
|
|
| ## 用法 |
|
|
| ### 直接运行 |
| ```bash |
| chmod +x bin/visdrone_detect |
| LD_LIBRARY_PATH=/soc/lib ./bin/visdrone_detect model.axmodel image.jpg 0.25 |
| ``` |
|
|
| ### 集成到自己的项目 |
| ```cpp |
| #include "libdet.h" |
| |
| ax_det_init_t init = {}; |
| init.dev_type = axcl_device; |
| init.model_type = ax_det_model_type_yolo11; |
| sprintf(init.model_path, "model.axmodel"); |
| init.num_classes = 10; |
| init.threshold = 0.25f; |
| |
| ax_det_handle_t handle; |
| ax_det_init(&init, &handle); |
| |
| // 加载 cv::Mat 图像 (RGB uint8), resize 到 640x640 |
| ax_det_img_t img = {.data = rgb_data, .width = 640, .height = 640, |
| .channels = 3, .stride = 640 * 3}; |
| ax_det_result_t result; |
| ax_det(handle, &img, &result); |
| |
| ax_det_deinit(handle); |
| ``` |
|
|
| 编译: |
| ```bash |
| aarch64-none-linux-gnu-g++ -std=c++17 \ |
| -I include -L lib -o my_detect my_detect.cpp \ |
| -ldet -lpthread -ldl |
| ``` |
|
|
| ## 编译(如需) |
| 源码和 CMakeLists.txt 见: https://github.com/ml-inory/visdrone-yolov11s.axera |
| 交叉编译依赖: |
| - AX650 BSP SDK V3.10.2(OpenCV 4.5.5 aarch64) |
| - aarch64-none-linux-gnu-g++ 9.2+ |
|
|