Ultra96 v2 v2.4 github repo problem

Hi, i am a student from Taiwan. I want to download ‘pynq computervision’ from github to the ultra96 v2 board with sdcard image v2.4
I use the instruction “sudo pip3 install --upgrade git” .
However, it just seem stopped at “Cloning to /tmp/pip-4k1rh8xx-build” .
I have seen lots of website and still do not know where it is wrong.
Hope someone can help me on it.
And my English is not very good,sorry for that.
Thanks a lot a lot a lot!!

It’s likely the problem is with the internet connection on the board. Are you able to clone the repository directly?

git clone

If so you can install the repository directly by going into the directory and running

sudo pip3 install .

Another thing you can do is add the -v flag to the pip install command which should produce more debugging information.


Thanks a lot!
I change my internet and it can download now,but i meet another problem.
Both sdcard with v2.4 and v2.5 meet this problem too.
Thank you for your helping!!


Try running the pip command without the --upgrade

sudo pip3 install git

It looks like there has been a new version of the pandas library which isn’t supported on our current platform.


It still meet the same problem. Is there any other repo about using opencv to deel with image or video on ultra96 v2 board? Thanks a lot!


I had the same problem.
How I fixed it:

pip3 install --upgrade Cython
sudo pip3 install --upgrade git+

Then you will probably run into this problem:
ImportError: cannot open shared object file: No such file or directory

Follow the solution from rock: ImportError: cannot open shared object file: No such file or directory

You have to manually install the libopencv packages (around 17) after that it should work :smile:

I used PYNQ 2.5 on an Ultra96-V2

1 Like

Thanks a lot! Awesome!
There is another problem i want to ask~
I do not why the fps of running HW filter2D + dilate with Computervision on input frames from USB camera in real-time is only 4~5 fps.
But the fps setting is about 30 fps. Is it the problem that jupyter is not good at running GUI?
Thanks a lot!

Hey @TingShen_Kuo . That is really interesting question. Actually I’ve run into a similar problem and searched a little but I didn’t try it out because I don’t have a webcam. But I would be interested in the results. I am not sure whether it is a hardware limitation of your camera or whether opencv limits the framerate or if jupyter isn’t supporting
a higher framerate. (I would guess it is a limitation of jupyer to display your video in the webbrowser). But maybe you can use the DisplayPort as an output for your Webcam. Here is an example but for the Pynq-Z1. Not sure if there is an already build notebook for the Ultra96-V2. If you find one please send me it :joy:

Maybe somebody else can help :thinking:

Here are some links for using videos in jupyter.
Here. Here1. Here2.