1
cd ~
2
wget --user-agent="Mozilla" https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda3-2024.10-1-Linux-aarch64.sh
3
sh Anaconda3-2024.10-1-Linux-aarch64.sh
4
# 如果不能使用conda命令，在环境变量最后加上
5
nano ~/.bashrc
6
export PATH=/home/orangepi/anaconda3/bin:$PATH
7
source ~/.bashrc

1
conda create -y -n yolo11 python=3.10
2
conda activate yolo11

1
# yolo11n
2
wget https://ftrg.zbox.filez.com/v2/delivery/data/95f00b0fc900458ba134f8b180b3f7a1/examples/yolo11/yolo11n.onnx
3
# yolo11s
4
wget https://ftrg.zbox.filez.com/v2/delivery/data/95f00b0fc900458ba134f8b180b3f7a1/examples/yolo11/yolo11s.onnx
5
# yolo11m
6
wget https://ftrg.zbox.filez.com/v2/delivery/data/95f00b0fc900458ba134f8b180b3f7a1/examples/yolo11/yolo11m.onnx

1
conda install -c conda-forge cmake=3.25.*

1
git clone --branch v2.3.0 https://github.com/airockchip/rknn-toolkit2.git
2
cd rknn-toolkit2/packages/arm64
3

4
conda activate yolo11
5
pip install -r arm64_requirements_cp310.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
6
pip install rknn_toolkit2-2.3.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
7

8
cd ~/rknn-toolkit2/rknn-toolkit-lite2/packages
9
pip install rknn_toolkit_lite2-2.3.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
10

11
# 把 RKNN-Toolkit2 里的可执行文件和动态库手动复制到系统的标准路径中，方便系统识别和调用
12
cd ~/rknn-toolkit2/rknpu2/runtime/Linux/rknn_server/aarch64/usr/bin
13
sudo cp * /usr/bin/
14
cd ~/rknn-toolkit2/rknpu2/runtime/Linux/librknn_api/aarch64
15
sudo cp * /usr/lib/

1
# 若版本不对，使用下面命令重新安装torch
2
pip install torch==1.13.1 -i https://pypi.tuna.tsinghua.edu.cn/simple

1
git clone https://github.com/airockchip/rknn_model_zoo.git

1
# 将ONNX模型放在rknn_model_zoo/examples/yolo11/model
2
cd rknn_model_zoo/examples/yolo11/python
3
python convert.py ../model/best.onnx rk3588 # model文件夹会得到yolo11.rknn

1
python yolo11.py --model_path ../model/yolo11.rknn --target rk3588 --img_show

1
cat /sys/kernel/debug/rknpu/driver_version

1
sudo watch -n 1 "cat /sys/kernel/debug/rknpu/load"