search

prs-eth / OverlapPredator

[CVPR 2021, Oral] PREDATOR: Registration of 3D Point Clouds with Low Overlap.
https://shengyuh.github.io/predator/index.html
MIT License
512 stars 73 forks source link

Inference does not get applied #61

Open Birkenpapier opened 1 year ago

Birkenpapier commented 1 year ago

Hello everyone,

thank you very much for your work! After setting up your project on a Win10 machine with a dedicated conda environment I got following result: grafik

Everything seems to work without problems and I've even got the same message in the terminal regarding numpy as you got. But the inference and the overlapping areas are not getting visualized. Why is that the case?

Thank you very much! Here is also my Cuda version aswell the other dependencies: grafik

matplotlib==3.3.3 numpy==1.19.4 torch == 1.8.0 torchvision==0.9.0 torchaudio==0.8.0 nibabel==3.2.1 tqdm==4.38.0 open3d==0.10.0.0 easydict==1.9 scipy==1.5.4 coloredlogs==15.0 PyYAML==5.4.1 scikit_learn==0.24.1 tensorboardX==2.1 vtk_visualizer==0.9.6 nibabel==3.2.1 h5py==3.2.1 coloredlogs==15.0 gitpython==3.1.17

Thank you very much for your help!

tangqi1012 commented 1 year ago

I've encountered the same issue but I don't know what the cause is.

Hellcat1005 commented 1 year ago

I've encountered the same issue but I don't know what the cause is.

lucienchia commented 9 months ago

their demo.py result transformation matrix is a 4X4 identity matrix suggesting that there is no transformation at all.... @ShengyuH may i know why is this so?

xl232lx commented 3 weeks ago

Hello, I have also encountered the situation of poor registration effect, may I ask if you have solved it here？