ERROR: Could not install packages due to an OSError: [Errno 13] Permission denied:

[bwajster]

After running the commandpip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl and waiting for more than 3 hours I got the error below: ~/mediapipe-bin$ pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl WARNING: pip is being invoked by an old script wrapper. This will fail in a future version of pip. Please see https://github.com/pypa/pip/issues/5599 for advice on fixing the underlying issue. To avoid this problem you can invoke Python with '-m pip' instead of running pip directly. Defaulting to user installation because normal site-packages is not writeable Processing ./v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl Requirement already satisfied: numpy==1.19.4 in /usr/local/lib/python3.6/dist-packages (from mediapipe==0.8.5-cuda102) (1.19.4) Requirement already satisfied: wheel in /usr/lib/python3/dist-packages (from mediapipe==0.8.5-cuda102) (0.30.0) Collecting opencv-contrib-python Downloading opencv-contrib-python-4.9.0.80.tar.gz (152.6 MB) |████████████████████████████████| 152.6 MB 6.5 kB/s
Installing build dependencies ... done Getting requirements to build wheel ... done Preparing metadata (pyproject.toml) ... done Collecting absl-py Downloading absl_py-1.4.0-py3-none-any.whl (126 kB) |████████████████████████████████| 126 kB 5.2 MB/s
Requirement already satisfied: protobuf>=3.11.4 in /usr/local/lib/python3.6/dist-packages (from mediapipe==0.8.5-cuda102) (3.19.6) Collecting attrs>=19.1.0 Downloading attrs-22.2.0-py3-none-any.whl (60 kB) |████████████████████████████████| 60 kB 2.0 MB/s
Requirement already satisfied: six in /usr/lib/python3/dist-packages (from mediapipe==0.8.5-cuda102) (1.11.0) Building wheels for collected packages: opencv-contrib-python Building wheel for opencv-contrib-python (pyproject.toml) ... done Created wheel for opencv-contrib-python: filename=opencv_contrib_python-4.9.0.80-cp36-cp36m-linux_aarch64.whl size=21413170 sha256=fe83dc36fd60abf1d280846d24af33f6fe4392f45cfa92223fa78d745e09af62 Stored in directory: /home/nikhil/.cache/pip/wheels/d3/15/23/dafe20985a737773ed2f139e6917ecabc74d64cf17cfef1907 Successfully built opencv-contrib-python Installing collected packages: opencv-contrib-python, attrs, absl-py, mediapipe ERROR: Could not install packages due to an OSError: [Errno 13] Permission denied: '/home/nikhil/.local/lib/python3.6' Check the permissions.

Please help me resolve this error.

[bwajster]

Do i need to use sudo or/and -H flag or the --user flag while running the command: pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl ?? Is this it or what am I missing? It seems that the error is related to the permissions of the /home/nikhil/.local/lib/python3.6 directory, which is the user-specific Python package installation directory.

Please help.

[bwajster]

When I tried the command pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl with a sudo I got the warning below: sudo pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl [sudo] password for nikhil: WARNING: The directory '/home/nikhil/.cache/pip' or its parent directory is not owned or is not writable by the current user. The cache has been disabled. Check the permissions and owner of that directory. If executing pip with sudo, you should use sudo's -H flag.

[bwajster]

So taking into consideration the above errors and warnings what's the right way to run the command pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl ?

[lanzani]

Hi! The fact that it takes hours is right, or at least is what I also experenced.

The command should be run with sudo:

sudo pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl

I also got the warning, but it don't stop the installation process, try to run the above command and ignore the warning.

[bwajster]

Using a -H flag helps avoid the warning: sudo -H pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl

[bwajster]

Does the last commandpip3 install dataclasses need a sudo too ?

[bwajster]

Also, are there other ways to check if MediaPipe uses the GPU? Apart from seeing log-Created TensorFlow Lite XNNPACK delegate for GPU printed in the terminal?

[lanzani]

Does the last commandpip3 install dataclasses need a sudo too ?

Try using sudo

[lanzani]

Also, are there other ways to check if MediaPipe uses the GPU? Apart from seeing log-Created TensorFlow Lite XNNPACK delegate for GPU printed in the terminal?

Do you get that message? If yes, congratulations! You have mediapipe with gpu :)

[lanzani]

You can test it using this script, or one of your own and see if the framerate is accettable.

With the script above, I got around 20 fps with a HD webcam.

[bwajster]

Ok, thank you.

[lanzani]

Let me know if it works for you :)

[bwajster]

Sure, I'll keep you posted.

[bwajster]

After running the command sudo -H pip3 install v0.8.5/v0.8.5/numpy119x/py36/mediapipe-0.8.5_cuda102-cp36-cp36m-linux_aarch64.whl here's what I got:

And after running the pose estimation script, here's what I got: On the pic above I've highlighted the line:Created TensorFlow Lite delegate for GPU. But, on your site to validate GPU support the log is different: Created TensorFlow Lite XNNPACK delegate for GPU.

Please confirm if 1. MediaPipe is installed successfully and 2. If the script runs with GPU support.

[lanzani]

According to your screenshots, you have mediapipe correctly installed, so:

1. Yes
2. Yes

On my website there is written: Created TensorFlow Lite XNNPACK delegate for GPU because maybe I copied and pasted the message from another version of mediapipe. But according to your screenshot, I can say that you installed everything successfully :)

[bwajster]

Great, a big thank you for having the instructions and steps explained so clearly on your website, I couldn't have done it without your efforts.

[lanzani]

Thank you very much! If you want to, leave a star to this repository :)