Closed zeyu659 closed 3 months ago
Hi @zeyu659
This is probably linked to how you have configured your all_base_cloud_ids
.
Our BaseDataset
class inherits from PyG's Dataset
class. As explained here. The download()
method is called when the Datatest
could not find all the raw_file_names
in the raw/
directory.
Have a closer look at the PyG documentation and our BaseDataset
code to understand how the raw_dir
and processed_dir
work. In particular, for your situation, I would suggest you have a closer look at how the following work:
all_base_cloud_ids
all_cloud_ids
cloud_ids
raw_file_names
dear @drprojects , I am having problems creating my own dataset 'scaffolds' and running train.py. I have read the create your own dataset section and created 'scaffolds' in the format of the ‘scannet’ dataset. Created _/src/datasets/scaffolds.py & scaffoldsconfig.py, /src/datamodules/scaffolds.py, configs/datamodule/ semantic/scaffolds.yaml, _configs/experiment/semantic/scaffolds11g.yaml. A new folder /data/scaffolds/raw/ is created to save the ‘scaffolds’ dataset in it, the ‘train&val’ part is saved in the /train folder as the scannet dataset is all saved in /scan folder, and the ‘test’ part is saved in the /test folder, the content of which is that the sem_label of each point has already been matched with the xyzrgb matching in the form of a sceneXXXX_point_clouds_all.npz file as follows:
However, when I try to run train.py, it will get the error as follow. I looked at my /data/scaffolds folder, which does have my own data already saved in it following the above structure, and I'm not sure what went wrong to cause it to jump straight to when the code is run:
By the way, /ssd1/gaozy/Code/superpoint_transformer/data/scannet also exists correctly and runs successfully. Thank you in advance for your help and support.