qt.qpa.screen: QXcbConnection: Could not connect to display

[2feet6inches]

$ ns-process-data --data room.mov --output-dir data/

Number of frames in video: 1616
[07:47:09] 🎉 Done converting video to images.                                                       process_data.py:433
[07:47:21] 🎉 Done downscaling images.                                                               process_data.py:451
──────────────────────────────────────────────  💀 💀 💀 ERROR 💀 💀 💀  ───────────────────────────────────────────────
Error running command: colmap feature_extractor --database_path data/colmap/database.db --image_path data/images
--ImageReader.single_camera 1 --ImageReader.camera_model OPENCV --SiftExtraction.use_gpu 1
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
qt.qpa.screen: QXcbConnection: Could not connect to display
Could not connect to any X display.

[2feet6inches]

Probably the same issue as this. I was able to run it by adding the --no-gpu flag. https://github.com/colmap/colmap/issues/18

[Wushuangpin]

应该是和这个一样的问题。我可以通过添加 --no-gpu 标志来运行它。 colmap/colmap#18 This is fine, but the preprocessing time is too long without gpu acceleration. Is there another solution: use gpu acceleration without opening the colmap graphics window.

[paudom]

Hi @Wushuangpin do you know if the solution of using GPU without opening graphics window is possible as the quoted message notes? How can be done? Thanks in advance!

[Zunhammer]

This issue (re)occured a few days ago to me in some testing environment which is most likely to changes in the dev branch of colmap. You should always use defined versions, optimally a release like 3.7 and you should be fine again. See Colmap v3.7

Can you please try this and give information about the colmap version you're using?

[tancik]

Closing as it appears to be stale. Feel free to reopen if needed.

[pengc02]

Since sudo apt install colmap usually installs without CUDA，conda install -c conda-forge colmap work for me.