sayands / sgaligner

[ICCV 2023] SGAligner : 3D Scene Alignment with Scene Graphs
https://sayands.github.io/sgaligner/
MIT License
81 stars 4 forks source link

Missing "raw points" #2

Closed EryiXie closed 1 year ago

EryiXie commented 1 year ago

Hi, I am trying to run SGAligner for registration, there is this line of code makes me confused, after generating the data samples as described in the README, I can not find the location of this "raw point" file. I know it should be the sampled point cloud of each scan, but I am not sure if I should change this line to read the raw .ply file from the original 3RScan or if it should be the downsampled point cloud of the raw .ply file. https://github.com/sayands/sgaligner/blob/49ae3e1398e369557878af7c45252817a7abe72f/src/inference/sgaligner/inference_align_reg.py#L144C21-L144C124 Thank you in advance.

sayands commented 1 year ago

Hi,

I am not exactly sure what you mean, the “raw” point cloud should be the point cloud file from 3RScan. Please note that we converted the .ply files to .npy files for faster processing.

EryiXie commented 1 year ago

Hi, thank you very much for the reply, now I understand. This lien of code loads the converted scans(point cloud) from 3RScan as the "raw points".

raw_points = scan3r.load_plydata_npy(osp.join(self.test_dataset.scans_scenes_dir, scan_id, 'data.npy'))

But It looks like after preprocessing 3RScan dataset using the scripts given in this repository, I obtain the "subscans_scenes_dir/scan_id/data.npy" file for each subscans, but not the "scan_scenes_dir/scan_id/data.npy".

But anyway, now I am clear what I should so is simply converting the ply file to npy file for each scans. Thank you again ^_^.

sguttikon commented 1 year ago

@sayands would be nice if you can add this information to README.md and/or update preprocessing code, since the downloaded 3RScan doesn't contain those files by default.

Hi,

I am not exactly sure what you mean, the “raw” point cloud should be the point cloud file from 3RScan. Please note that we converted the .ply files to .npy files for faster processing.