Skip to content

Latest commit

 

History

History
115 lines (103 loc) · 3.86 KB

DATASETS.md

File metadata and controls

115 lines (103 loc) · 3.86 KB
Raw

Datasets

All datasets go into a data directory at the root of the project. You can use for example mkdir data or ln -s /my/data/storage data to create this directory.

In the end the overall directory structure should be as follows (generated with tree data -L 2).

data
├── modelnet40_ply_hdf5_2048
│   ├── ply_data_test0.h5
│   ├── ply_data_test_0_id2file.json
│   ├── ply_data_test1.h5
│   ├── ply_data_test_1_id2file.json
│   ├── ply_data_train0.h5
│   ├── ply_data_train_0_id2file.json
│   ├── ply_data_train1.h5
│   ├── ply_data_train_1_id2file.json
│   ├── ply_data_train2.h5
│   ├── ply_data_train_2_id2file.json
│   ├── ply_data_train3.h5
│   ├── ply_data_train_3_id2file.json
│   ├── ply_data_train4.h5
│   ├── ply_data_train_4_id2file.json
│   ├── shape_names.txt
│   ├── test_files.txt
│   └── train_files.txt
├── ModelNetFewshot
│   ├── 10way_10shot
│   ├── 10way_20shot
│   ├── 5way_10shot
│   └── 5way_20shot
├── ScanObjectNN
│   ├── main_split
│   ├── main_split_nobg
│   ├── split1
│   ├── split1_nobg
│   ├── split2
│   ├── split2_nobg
│   ├── split3
│   ├── split3_nobg
│   ├── split4
│   └── split4_nobg
├── ShapeNet55
│   ├── shapenet_pc
│   ├── shapenet_test.npz
│   ├── shapenet_train.npz
│   ├── test.txt
│   ├── train_25.txt
│   ├── train_50.txt
│   ├── train_75.txt
│   └── train.txt
└── shapenetcore_partanno_segmentation_benchmark_v0_normal
    ├── 02691156
    ├── 02773838
    ├── 02954340
    ├── 02958343
    ├── 03001627
    ├── 03261776
    ├── 03467517
    ├── 03624134
    ├── 03636649
    ├── 03642806
    ├── 03790512
    ├── 03797390
    ├── 03948459
    ├── 04099429
    ├── 04225987
    ├── 04379243
    ├── synsetoffset2category.txt
    └── train_test_split

ShapeNet

Download ShapeNet55.zip from Google Drive into the data directory (see also Point-BERT DATASET.md). Then

cd data
unzip ShapeNet55.zip
cp ../.metadata/ShapeNet55/* ShapeNet55/ # train.txt and test.txt split files
cd ..
python -m point2vec.datasets.process.shapenet_npz

The last step saves the whole dataset in shapenet_train.npz and shapenet_test.npz files which are used by the --data.in_memory true flag. This is recommended if you have slow disk I/O, and necessary if you want to reproduce the exact results from our paper.

ScanObjectNN

Download h5_files.zip from the official Website into the data directory by agreeing to the Terms of Use. Then

cd data
unzip h5_files.zip
mv h5_files ScanObjectNN

ModelNet40

cd data
wget --no-check-certificate https://shapenet.cs.stanford.edu/media/modelnet40_ply_hdf5_2048.zip
unzip modelnet40_ply_hdf5_2048.zip

ModelNet Few Shot

Download the directory from Google Drive into the data directory (see also Point-BERT DATASET.md).

ShapeNetPart

cd data
wget --no-check-certificate https://shapenet.cs.stanford.edu/media/shapenetcore_partanno_segmentation_benchmark_v0_normal.zip
unzip shapenetcore_partanno_segmentation_benchmark_v0_normal.zip