YOLO environment construction

1. System information

2. Preliminary preparation

sudo apt update
sudo apt install python3-pip -y
sudo pip install -U pip

3. Install Ultralytics

xxxxxxxxxx
sudo pip3 install ultralytics[export]

xxxxxxxxxx
sudo reboot

4. Configure GPU acceleration

torch

xxxxxxxxxx
sudo pip3 install https://github.com/ultralytics/assets/releases/download/v0.0.0/torch-2.5.0a0+872d972e41.nv24.08-cp310-cp310-linux_aarch64.whl

torchvision

xxxxxxxxxx
sudo pip3 install https://github.com/ultralytics/assets/releases/download/v0.0.0/torchvision-0.20.0a0+afc54f7-cp310-cp310-linux_aarch64.whl

cuSPARSELt

xxxxxxxxxx
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/arm64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install libcusparselt0 libcusparselt-dev

onnxruntime-gpu

xxxxxxxxxx
sudo pip3 install https://github.com/ultralytics/assets/releases/download/v0.0.0/onnxruntime_gpu-1.20.0-cp310-cp310-linux_aarch64.whl

Note: Using onnxruntime-gpu requires installing a specific version of numpy. If it is not 1.23.5, you can run the following command to install the specified version

xxxxxxxxxx
sudo pip3 install numpy==1.23.5

5. Verify the installation

Verifying Ultralytics

xxxxxxxxxx
python3 -c "import ultralytics; print(ultralytics.__version__)"

Verifying Torch

xxxxxxxxxx
python3 -c "import torch; print(torch.__version__); print(torch.cuda.is_available())"

Verifying Torchvision

xxxxxxxxxx
python3 -c "import torchvision; print(torchvision.__version__)"

Verify Numpy

xxxxxxxxxx
python3 -c "import numpy; print(numpy.__version__)"

Note: The ultralytics version will be updated later, please refer to the version information queried in the system

Common Errors

Cannot uninstall sympy

Error phenomenon

Unable to uninstall sympy

Solution

Uninstall python3-sympy: Reinstall PyTorch after uninstallation

xxxxxxxxxx
sudo apt remove python3-sympy -y

CSI camera cannot be called

Compile OpenCV from source code and enable GStreamer support: basically the entire process is automatically installed. It is recommended to uninstall the old version and install the new version (the script automatically enables CUDA and GStreamer functions)

xxxxxxxxxx
git clone https://github.com/AastaNV/JEP.git
cd JEP/script
bash install_opencv4.10.0_Jetpack6.1.sh

Verify the environment

xxxxxxxxxx
python3 -c "import cv2; print(cv2.getBuildInformation())" | grep GStreamer
python3 -c "import ultralytics; print(ultralytics.__version__)"
python3 -c "import torch; print(torch.__version__); print(torch.cuda.is_available())"
python3 -c "import torchvision; print(torchvision.__version__)"
python3 -c "import numpy; print(numpy.__version__)"
jtop

References

https://docs.ultralytics.com/guides/nvidia-jetson/

https://github.com/AastaNV/JEP