PP-YOLOE

环境配置

信息

参考 RKNN 安装 配置好相关环境。

参考 RKNN Model Zoo 下载示例文件。

模型下载

下载 onnx 模型文件。

X64 Linux PC

cd rknn_model_zoo/examples/ppyoloe/model/
bash download_model.sh

模型转换

选择目标平台。

rk3588

X64 Linux PC

export TARGET_PLATFORM=rk3588

rk356x

X64 Linux PC

export TARGET_PLATFORM=rk356x

rk3576

X64 Linux PC

export TARGET_PLATFORM=rk3576

将 onnx 模型转换为 rknn 模型。

X64 Linux PC

cd ../python/
python convert.py ../model/ppyoloe_s.onnx ${TARGET_PLATFORM}

C API

编译示例

切换到 rknn_model_zoo 目录下执行 build-linux.sh 编译脚本。

X64 Linux PC

cd ../../..
bash build-linux.sh -t ${TARGET_PLATFORM} -a aarch64 -d ppyoloe

文件同步

然后将编译生成的 install 目录下的 demo 目录推送到板端。

X64 Linux PC

cd install/${TARGET_PLATFORM}_linux_aarch64/
scp -r rknn_ppyoloe_demo/ user@your_device_ip:target_directory

运行示例

导出运行时库到环境变量。

Device

cd rknn_ppyoloe_demo/
export LD_LIBRARY_PATH=./lib

运行示例。

Device

./rknn_ppyoloe_demo ./model/ppyoloe.rknn ./model/bus.jpg

$ ./rknn_ppyoloe_demo ./model/ppyoloe.rknn ./model/bus.jpg
load label ./model/coco_80_labels_list.txt
model input num: 1, output num: 9
input tensors:
  index=0, name=image, n_dims=4, dims=[1, 640, 640, 3], n_elems=1228800, size=1228800, fmt=NHWC, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
output tensors:
  index=0, name=conv2d_176.tmp_1, n_dims=4, dims=[1, 68, 20, 20], n_elems=27200, size=27200, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-42, scale=0.072882
  index=1, name=sigmoid_2.tmp_0, n_dims=4, dims=[1, 80, 20, 20], n_elems=32000, size=32000, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003802
  index=2, name=clip_0.tmp_0, n_dims=4, dims=[1, 1, 20, 20], n_elems=400, size=400, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
  index=3, name=conv2d_182.tmp_1, n_dims=4, dims=[1, 68, 40, 40], n_elems=108800, size=108800, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-47, scale=0.085614
  index=4, name=sigmoid_5.tmp_0, n_dims=4, dims=[1, 80, 40, 40], n_elems=128000, size=128000, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003736
  index=5, name=clip_1.tmp_0, n_dims=4, dims=[1, 1, 40, 40], n_elems=1600, size=1600, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
  index=6, name=conv2d_188.tmp_1, n_dims=4, dims=[1, 68, 80, 80], n_elems=435200, size=435200, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-43, scale=0.102416
  index=7, name=sigmoid_8.tmp_0, n_dims=4, dims=[1, 80, 80, 80], n_elems=512000, size=512000, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003279
  index=8, name=clip_2.tmp_0, n_dims=4, dims=[1, 1, 80, 80], n_elems=6400, size=6400, fmt=NCHW, type=INT8, qnt_type=AFFINE, zp=-128, scale=0.003922
model is NHWC input fmt
model input height=640, width=640, channel=3
origin size=640x640 crop size=640x640
input image: 640 x 640, subsampling: 4:2:0, colorspace: YCbCr, orientation: 1
scale=1.000000 dst_box=(0 0 639 639) allow_slight_change=1 _left_offset=0 _top_offset=0 padding_w=0 padding_h=0
rga_api version 1.10.1_[0]
rknn_run
person @ (108 232 224 536) 0.943
person @ (478 232 561 519) 0.928
person @ (211 240 283 512) 0.916
bus @ (88 135 552 442) 0.909
person @ (78 326 125 516) 0.527
handbag @ (261 339 281 413) 0.420
handbag @ (253 342 264 380) 0.272
write_image path: out.png width=640 height=640 channel=3 data=0x6b49a30

效果展示

Python API

激活虚拟环境

Device

conda activate rknn

运行示例

将相关文件推送到板端执行下面的命令。

Device

python ppyoloe.py --model_path ./model/ppyoloe.rknn --target ${TARGET_PLATFORM} --img_folder ./model/ --img_save

$ python ppyoloe.py --model_path ./model/ppyoloe.rknn --target rk3588 --img_folder ./model/ --img_save
/home/radxa/miniforge3/envs/rknn/lib/python3.12/site-packages/rknn/api/rknn.py:51: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools<81.
  self.rknn_base = RKNNBase(cur_path, verbose)
I rknn-toolkit2 version: 2.3.2
--> Init runtime environment
I target set by user is: rk3588
done
Model-./model/ppyoloe.rknn is rknn model, starting val
W inference: The 'data_format' is not set, and its default value is 'nhwc'!


IMG: bus.jpg
class: person, score: 0.942828357219696
box coordinate left,top,right,down: [108, 232, 224, 536]
class: person, score: 0.9276214241981506
box coordinate left,top,right,down: [478, 232, 561, 519]
class: person, score: 0.916216254234314
box coordinate left,top,right,down: [211, 240, 283, 512]
class: person, score: 0.5267177820205688
box coordinate left,top,right,down: [78, 326, 125, 516]
class: handbag, score: 0.41975846886634827
box coordinate left,top,right,down: [261, 339, 281, 413]
class: handbag, score: 0.27218714356422424
box coordinate left,top,right,down: [253, 342, 264, 380]
class: bus , score: 0.9086127877235413
box coordinate left,top,right,down: [88, 135, 552, 442]
Detection result save to ./result/bus.jpg

效果展示