scanobjectnn_PB_T50_RS_h5.zip contains h5 files for the hard variant of the ScanObjectNN benchmark.
Dataset can be loaded as follows:
import os.path as osp
import os
import torch
import h5py
import torch_geometric.transforms as T
from torch_geometric.datasets import ModelNet
from torch_geometric.data import InMemoryDataset, download_url, extract_zip, Data
class ScanObjectNN(InMemoryDataset):
url =… See the full description on the dataset page:
https://huggingface.co/datasets/cminst/ScanObjectNN.
