RuyiSDK AI 示例
安装依赖库
sudo apt update
sudo apt install git make build-essential安装 RuyiSDK
# 下载并安装 ruyi
wget https://mirror.iscas.ac.cn/ruyisdk/ruyi/tags/0.47.0/ruyi-0.47.0.amd64
chmod +x ruyi-0.47.0.amd64
sudo cp ruyi-0.47.0.amd64 /usr/local/bin/ruyi安装工具链
#初始化并更新 Ruyi 仓库
ruyi update
#下载 Duo 256M 专属工具链
ruyi install gnu-milkv-milkv-duo-musl-binYOLOv5
示例描述和硬件环境准备
基于 TDL SDK 实现 YOLOv5 目标检测，读取图像并进行推理，验证 Ruyi 工具链在 Duo256M（RISC-V）上的交叉编译能力。
硬件环境：Milk-V Duo256M （SG2002 ）
软件环境：Ubuntu，RuyiSDK 0.47.0，TDL SDK
创建并激活 ruyi 虚拟环境
# 创建虚拟环境
ruyi venv -t gnu-milkv-milkv-duo-musl-bin generic milkv-venv
# 激活
source milkv-venv/bin/ruyi-activate
# 验证
which riscv64-unknown-linux-musl-gcc在终端显示如下：
«Ruyi milkv-venv» duo256@duo256-virtual-machine:~$ which riscv64-unknown-linux-musl-gcc
/home/duo256/milkv-venv/bin/riscv64-unknown-linux-musl-gcc使用 ruyi 工具链编译示例代码
下载 TDL SDK 源码
git clone https://github.com/milkv-duo/cvitek-tdl-sdk-sg200x.git #进入代码目录进行编译，生成sample_yolov5 程序
cd cvitek-tdl-sdk-sg200x/sample/cvi_yolo
make sample_yolov5 \
KERNEL_ROOT=$HOME/cvitek-tdl-sdk-sg200x/sample \
MW_PATH=$HOME/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2 \
TPU_PATH=$HOME/cvitek-tdl-sdk-sg200x/sample/3rd/tpu \
IVE_PATH=$HOME/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2 \
CHIP=CV181X \
SDK_VER=musl_riscv64在终端显示如下：
«Ruyi milkv-venv» duo256@duo256-virtual-machine:~$ cd cvitek-tdl-sdk-sg200x/sample/cvi_yolo
«Ruyi milkv-venv» duo256@duo256-virtual-machine:~/cvitek-tdl-sdk-sg200x/sample/cvi_yolo$ make sample_yolov5 \
KERNEL_ROOT=$HOME/cvitek-tdl-sdk-sg200x/sample \
MW_PATH=$HOME/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2 \
TPU_PATH=$HOME/cvitek-tdl-sdk-sg200x/sample/3rd/tpu \
IVE_PATH=$HOME/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2 \
CHIP=CV181X \
SDK_VER=musl_riscv64
---------------------------------------
CHIP: CV181X
SDK_VER: musl_riscv64
TDL SDK library path: /home/duo256/cvitek-tdl-sdk-sg200x/lib
TDL SDK include path: /home/duo256/cvitek-tdl-sdk-sg200x/include
Middleware include path: /home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include
Middleware library path: /home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib
IVE library path: /home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib
IVE include path: /home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include
TPU library path: /home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/tpu/lib
CFLAGS: -mcpu=c906fdv -march=rv64imafdcv0p7xthead -mcmodel=medany -mabi=lp64d -O3 -DNDEBUG -D_MIDDLEWARE_V2_ -Wno-maybe-uninitialized -DCV181X -std=gnu11 -Wno-pointer-to-int-cast -fsigned-char -Wno-format-truncation -fdiagnostics-color=always -s -lpthread -latomic -I/home/duo256/cvitek-tdl-sdk-sg200x/include -I/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl -I/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl_app -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/utils -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/rtsp/include/cvi_rtsp -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/opencv/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/stb/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/sample/common -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/isp/cv181x/ -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/panel -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/linux -I/include/stb
CC: riscv64-unknown-linux-musl-gcc
CXX: riscv64-unknown-linux-musl-g++
USE_TPU_IVE:
SAMPLW_MW_LIB: -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib/3rd -lini -lsns_full -lsample -lisp -lvdec -lvenc -lawb -lae -laf -lcvi_bin -lcvi_bin_isp -lmisc -lisp_algo -lsys -lvpu -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/lib
SAMPLE_AUD_LIBS: -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib/3rd -lini -lsns_full -lsample -lisp -lvdec -lvenc -lawb -lae -laf -lcvi_bin -lcvi_bin_isp -lmisc -lisp_algo -lsys -lvpu -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/lib -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/tpu/lib -lcnpy -lcvikernel -lcvimath -lcviruntime -lz -lm -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib -lcvi_ive -L/home/duo256/cvitek-tdl-sdk-sg200x/lib -lcvi_tdl -lcvi_audio -lcvi_vqe -lcvi_ssp -ltinyalsa -lcvi_VoiceEngine -lcvi_RES1 -lcli
---------------------------------------
riscv64-unknown-linux-musl-g++ -mcpu=c906fdv -march=rv64imafdcv0p7xthead -mcmodel=medany -mabi=lp64d -O3 -DNDEBUG -D_MIDDLEWARE_V2_ -Wno-maybe-uninitialized -DCV181X -std=gnu11 -Wno-pointer-to-int-cast -fsigned-char -Wno-format-truncation -fdiagnostics-color=always -s -lpthread -latomic -I/home/duo256/cvitek-tdl-sdk-sg200x/include -I/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl -I/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl_app -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/utils -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/rtsp/include/cvi_rtsp -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/opencv/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/stb/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/sample/common -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/isp/cv181x/ -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/panel -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/linux -I/include/stb -o /home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.o -c /home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.cpp
cc1plus: warning: command-line option '-Wno-pointer-to-int-cast' is valid for C/ObjC but not for C++
cc1plus: warning: command-line option '-std=gnu11' is valid for C/ObjC but not for C++
/home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.cpp:1: warning: "_GNU_SOURCE" redefined
1 | #define _GNU_SOURCE
|
<command-line>: note: this is the location of the previous definition
In file included from /home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/cvi_tdl.h:5,
from /home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.cpp:14:
/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/core/cvi_tdl_core.h:173:79: warning: backslash and newline separated by space
173 | L_SUPPORTED_MODEL_YOLOV8_HARDHAT) \
|
In file included from /home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/core/cvi_tdl_custom.h:3,
from /home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/cvi_tdl.h:6,
from /home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.cpp:14:
/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/core/cvi_tdl_core.h:173:79: warning: backslash and newline separated by space
173 | L_SUPPORTED_MODEL_YOLOV8_HARDHAT) \
|
In file included from /home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/core/cvi_tdl_utils.h:3,
from /home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/cvi_tdl.h:7,
from /home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.cpp:14:
/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl/core/cvi_tdl_core.h:173:79: warning: backslash and newline separated by space
173 | L_SUPPORTED_MODEL_YOLOV8_HARDHAT) \
|
riscv64-unknown-linux-musl-g++ -mcpu=c906fdv -march=rv64imafdcv0p7xthead -mcmodel=medany -mabi=lp64d -O3 -DNDEBUG -D_MIDDLEWARE_V2_ -Wno-maybe-uninitialized -DCV181X -std=gnu11 -Wno-pointer-to-int-cast -fsigned-char -Wno-format-truncation -fdiagnostics-color=always -s -lpthread -latomic -I/home/duo256/cvitek-tdl-sdk-sg200x/include -I/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl -I/home/duo256/cvitek-tdl-sdk-sg200x/include/cvi_tdl_app -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/utils -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/rtsp/include/cvi_rtsp -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/opencv/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/stb/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/sample/common -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/isp/cv181x/ -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/panel -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include -I/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/include/linux -I/include/stb -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib/3rd -lini -lsns_full -lsample -lisp -lvdec -lvenc -lawb -lae -laf -lcvi_bin -lcvi_bin_isp -lmisc -lisp_algo -lsys -lvpu -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/lib -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/tpu/lib -lcnpy -lcvikernel -lcvimath -lcviruntime -lz -lm -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/middleware/v2/lib -lcvi_ive -L/home/duo256/cvitek-tdl-sdk-sg200x/lib -lcvi_tdl -L/home/duo256/cvitek-tdl-sdk-sg200x/sample/3rd/opencv/lib -lopencv_core -lopencv_imgproc -lopencv_imgcodecs -o sample_yolov5 /home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.o
rm /home/duo256/cvitek-tdl-sdk-sg200x/sample/cvi_yolo/sample_yolov5.o模型准备
可以直接点击下载本文文档预编译好的yolov5_cv181x_int8_sym.cvimodel
或者参考 TPU-MLIR 文档自己生成。
测试图片准备
从coco2017选择一张图片作为测试图片，复制到当前文件夹：
cp dataset/COCO2017/val2017/000000000632.jpg .运行示例并验证结果
将编译好的 sample_yolov5、cvimodel、要推理的 jpg 图片，拷贝到板端然后执行二进制程序：：
scp sample_yolov5 yolov5_cv181x_int8_sym.cvimodel 000000000632.jpg root@192.168.42.1:/root/在终端显示如下：
duo256@duo256-virtual-machine:~/yolov5$ scp sample_yolov5 yolov5_cv181x_int8_sym.cvimodel 000000000632.jpg root@192.168.42.1:/root/
root@192.168.42.1's password:
sample_yolov5 100% 14KB 1.3MB/s 00:00
yolov5_cv181x_int8_sym.cvim 100% 8005KB 3.1MB/s 00:02
000000000632.jpg 100% 152KB 3.1MB/s 00:00
duo256@duo256-virtual-machine:~/yolov5$ 在板端上执行以下命令：
export LD_LIBRARY_PATH='/mnt/system/lib'
./sample_yolov5 ./yolov5_cv181x_int8_sym.cvimodel 000000000632.jpg在终端显示如下：
[root@milkv-duo]~# ./sample_yolov5 ./yolov5_cv181x_int8_sym.cvimodel 000000000632.jpg
version: 1.4.0
yolov5s Build at 2026-05-28 10:19:59 For platform cv181x
Max SharedMem size:5734400
model opened:./yolov5_cv181x_int8_sym.cvimodel
detect res: 338.046265 215.812744 427.360474 347.613159 0.897093 58
detect res: 0.000000 292.178070 408.717834 478.821899 0.871600 59
detect res: 183.598511 120.470505 240.307343 234.079102 0.806835 58
detect res: 253.451004 232.237274 342.765228 314.312286 0.742620 56