Skip to content

kekxv/luckfox-bazel

BranchesTags

Repository files navigation

中文 English

Luckfox Bazel 项目介绍

本项目旨在为 Luckfox (幸狐) 开发板提供一套高效的开发框架。

项目采用 Bazel 进行构建管理，集成了 OpenCV Mobile、RKNN 以及 RGA 等核心组件，帮助开发者快速搭建环境并上手开发。

1. 环境与构建

本项目基于 Linux 环境开发。推荐使用 Linux 原生系统、虚拟机 或 Windows WSL。

注意：如果在非 Linux 环境下开发，建议使用远程编译模式，详见 remote-docker-deploy/buildbarn/README.md。

项目已预定义 .bazelrc 配置文件，支持跨平台构建：

# .bazelrc 配置片段
build:linux                --platforms=@cc_toolchains_linux//:linux-x86_64
build:linux-luckfox        --platforms=@cc_toolchains_linux//:linux-armv7l-luckfox

# 额外执行平台配置
build:linux-remote         --host_platform=@cc_toolchains_linux//:linux-x86_64
build:linux-remote         --extra_execution_platforms=@cc_toolchains_linux//:linux-x86_64

编译示例

只需指定 --config=linux-luckfox，即可编译出适用于 Luckfox 的目标产物：

  luckfox-bazel git:(main) $ bazel build --config=linux-luckfox test/helloworld:helloworld
...
INFO: Analyzed target //test/helloworld:helloworld (111 packages loaded, 3523 targets configured).
Target //test/helloworld:helloworld up-to-date:
  bazel-bin/test/helloworld/helloworld
INFO: Build completed successfully, 6 total actions

# 验证文件格式
  luckfox-bazel git:(main) $ file bazel-bin/test/helloworld/helloworld
bazel-bin/test/helloworld/helloworld: ELF 32-bit LSB executable, ARM, EABI5 version 1 (SYSV)...

2. 第三方依赖管理

为了简化开发流程，项目已将 librknn、librga、ncnn 及 opencv 等常用库封装为 Bazel 依赖：

third-party
├── librga
├── librknn
├── ncnn
└── zxing-cpp

接入方式

在 BUILD.bazel 中，只需将依赖添加到 deps 字段即可。

示例：集成 RKNN test/yolov5/BUILD.bazel

cc_library(
    name = "yolov5s",
    srcs = ["src/postprocess.cpp"],
    hdrs = ["include/postprocess.h"],
    defines = ["RV1106_RV1103"],
    deps = [
        ":yolov5s_640_640_rknn_bin",
        "//third-party/librknn",  # 引入 RKNN 依赖
    ],
)

示例：集成 RGA 与 OpenCV third-party/librga/BUILD.bazel

cc_binary(
    name = "test_rga",
    srcs = ["test/test.cc"],
    deps = [
        ":librga",
        "//third-party/librknn",
        "@opencv-mobile-luckfox//:opencv", # 引入 OpenCV 依赖
    ],
)

3. 实战案例：YOLOv5 推理

本节以 YOLOv5 为例，演示如何从源码编译到板端运行。

目录结构

test/yolov5
├── include          # 头文件
├── model            # 模型与标签 (rknn, txt)
├── src              # 后处理逻辑
└── test             # 主程序入口

编译与构建

直接通过 Bazel 构建目标：

  luckfox-bazel git:(main) $ bazel build --config=linux-luckfox test/yolov5:test-yolov5s  
...
INFO: Analyzed target //test/yolov5:test-yolov5s (40 packages loaded, 5542 targets configured).
INFO: From Generating C/C++ resources: yolov5s-640-640.rknn -> model_yolov5s-640-640...
Target //test/yolov5:test-yolov5s up-to-date:
  bazel-bin/test/yolov5/test-yolov5s
INFO: Build completed successfully, 18 total actions

板端运行结果

将生成的 test-yolov5s 可执行文件传输至 Luckfox 开发板并执行：

[root@luckfox tmp]# ./test-yolov5s bus.jpg
rknn_api/rknnrt version: 2.3.2 ...
model input num: 1, output num: 3
...
Begin perf ...
   0: Elapse Time = 100.09ms, FPS = 9.99
model is NHWC input fmt
 post_process Time = 10.06ms, FPS = 99.38
person @ (208 244 286 506) 0.884136
person @ (479 238 560 526) 0.863766
bus @ (94 130 553 464) 0.697389

从日志可以看出，模型成功加载并准确识别了图像中的目标，性能表现符合预期。

参考资料

Introduction to Luckfox Bazel Project

This project aims to provide an efficient development framework for the Luckfox development board.

It utilizes Bazel for build management and integrates core components such as OpenCV Mobile, RKNN, and RGA, allowing developers to quickly set up their environment and get started.

1. Environment & Build

This project is designed for a Linux environment. Native Linux, Virtual Machines (VMs), or Windows WSL are recommended.

Note: If developing in a non-Linux environment, it is recommended to use the remote build mode. Please refer to remote-docker-deploy/buildbarn/README.md for details.

The project comes with a pre-configured .bazelrc supporting cross-platform builds:

# .bazelrc snippet
build:linux                --platforms=@cc_toolchains_linux//:linux-x86_64
build:linux-luckfox        --platforms=@cc_toolchains_linux//:linux-armv7l-luckfox

# Extra execution platforms
build:linux-remote         --host_platform=@cc_toolchains_linux//:linux-x86_64
build:linux-remote         --extra_execution_platforms=@cc_toolchains_linux//:linux-x86_64

Build Example

Simply specify the --config=linux-luckfox flag to compile the target artifacts for Luckfox:

  luckfox-bazel git:(main) $ bazel build --config=linux-luckfox test/helloworld:helloworld
...
INFO: Analyzed target //test/helloworld:helloworld (111 packages loaded, 3523 targets configured).
Target //test/helloworld:helloworld up-to-date:
  bazel-bin/test/helloworld/helloworld
INFO: Build completed successfully, 6 total actions

# Verify file format
  luckfox-bazel git:(main) $ file bazel-bin/test/helloworld/helloworld
bazel-bin/test/helloworld/helloworld: ELF 32-bit LSB executable, ARM, EABI5 version 1 (SYSV)...

2. Third-party Dependencies

To simplify the development workflow, common libraries such as librknn, librga, ncnn, and opencv have been encapsulated as Bazel dependencies:

third-party
├── librga
├── librknn
├── ncnn
└── zxing-cpp

Integration

In BUILD.bazel, simply add the libraries to the deps field.

Example: RKNN Integration test/yolov5/BUILD.bazel

cc_library(
    name = "yolov5s",
    srcs = ["src/postprocess.cpp"],
    hdrs = ["include/postprocess.h"],
    defines = ["RV1106_RV1103"],
    deps = [
        ":yolov5s_640_640_rknn_bin",
        "//third-party/librknn",  # Import RKNN dependency
    ],
)

Example: RGA & OpenCV Integration third-party/librga/BUILD.bazel

cc_binary(
    name = "test_rga",
    srcs = ["test/test.cc"],
    deps = [
        ":librga",
        "//third-party/librknn",
        "@opencv-mobile-luckfox//:opencv", # Import OpenCV dependency
    ],
)

3. Practical Case: YOLOv5 Inference

This section demonstrates how to compile from source and run on the board using YOLOv5 as an example.

Directory Structure

test/yolov5
├── include          # Headers
├── model            # Models & Labels (rknn, txt)
├── src              # Post-processing logic
└── test             # Main entry point

Compilation

Build the target directly using Bazel:

  luckfox-bazel git:(main) $ bazel build --config=linux-luckfox test/yolov5:test-yolov5s  
...
INFO: Analyzed target //test/yolov5:test-yolov5s (40 packages loaded, 5542 targets configured).
INFO: From Generating C/C++ resources: yolov5s-640-640.rknn -> model_yolov5s-640-640...
Target //test/yolov5:test-yolov5s up-to-date:
  bazel-bin/test/yolov5/test-yolov5s
INFO: Build completed successfully, 18 total actions

Running on Board

Transfer the generated test-yolov5s executable to the Luckfox board and execute it:

[root@luckfox tmp]# ./test-yolov5s bus.jpg
rknn_api/rknnrt version: 2.3.2 ...
model input num: 1, output num: 3
...
Begin perf ...
   0: Elapse Time = 100.09ms, FPS = 9.99
model is NHWC input fmt
 post_process Time = 10.06ms, FPS = 99.38
person @ (208 244 286 506) 0.884136
person @ (479 238 560 526) 0.863766
bus @ (94 130 553 464) 0.697389

As seen in the logs, the model loads successfully and accurately identifies objects in the image, with performance meeting expectations.

References

About

luckfox on bazel

Resources

Readme
Activity

Stars

5 stars

Watchers

1 watching

Forks

1 fork
Report repository

Releases 4

luckfox on bazel Latest
Jan 20, 2026
+ 3 releases

Packages

 
 
 

Contributors

Languages