Matplotlib is currently using agg, which is a non-GUI backend, so cannot show the figure.

[zplvzrm]

Dear authors:

I am going to test the model effect on the 22.04 version of Ubuntu server with two NVIDIA GeForce RTX 3090 graphics cards. The docker basic environment I built is cuda11.1.1-cudnn8-devel-ubuntu18.04. It went smoothly in the early stage, but when I In the docker environment, when preparing to visualize the results of initial_pseudo_labels.pkl, I encountered error:

MoTTY X11 proxy: Unsupported authorisation protocol visualize_bev.py:444: UserWarning: Matplotlib is currently using agg, which is a non-GUI backend, so cannot show the figure. plt.show()

I followed the suggestions in visualize_bev.py and executed the command: apt-get update && apt-get install python3-tk But it has no effect, the error still exists. I also tried using tkagg, but it still didn't work. I hope you can give me some suggestions, I would be very grateful~

[darrenjkt]

Are you doing this: local computer > ssh into ubuntu server with RTX 3090 > run MS3D docker container? Then you're trying to run visualize_bev.py from the ssh terminal and expect a visualization window to open on your local computer? I assume this your use of ssh might be related to the MoTTY X11 proxy error.

If this is what you're doing, a potential fix could be using something like Teamviewer, VNCViewer or Remmina to do a remote desktop so that you can view the screen of your server. Then running the code in a terminal within that remote desktop session and it should open up a visualization window. If you want a cleaner solution, have a look at X11 forwarding for ssh.

For the agg error, maybe try pip install pyqt5. But I suspect that if you don't solve the MoTTY issue, the agg error wouldn't be solved either.