Skip to content

data.md

Latest commit

 

History

History
148 lines (118 loc) · 4.21 KB

data.md

File metadata and controls

148 lines (118 loc) · 4.21 KB

Data used for training and testing

Some datasets used for training are very large. If you do not want to retrain the model, you can skip them.

PF-Pascal (training and testing)

  1. Follow the instructions here to download and prepare the dataset. The structure should be:
PF-dataset-PASCAL
├──Annotations
├──JPEGImages
├── train_pairs_pf_pascal.csv
├── val_pairs_pf_pascal.csv
└── test_pairs_pf_pascal.csv

PF-Willow (testing only)

  1. Download PF-Willow dataset from the link.
  2. Rename the outermost directory from PF-dataset to pf-willow.
  3. Download lists for image pairs from link.
  4. Place the lists for image pairs under pf-willow directory. The structure should be:
pf-willow
├── PF-dataset
└── test_pairs.csv

SPair-71K (training and testing)

  1. Download SPair-71K dataset from link. After extraction, No more action required. The structure should be:
SPair-71k
├── ImageAnnotation
├── JPEGImages
└── PairAnnotation
└──...

AP-10k (training only)

  1. Download the AP-10k dataset and extract it.
./data/prepare_ap10k.sh
  1. Follow the instructions provided by GeoAware-SC at here to process AP-10k dataset. After processing, the structure should be:
ap-10k
├── annotations
├──── ap10k-train-split1.json
├──── ap10k-train-split2.json
├──── ap10k-train-split3.json
├──── ...
├── ImageAnnotation
├── JPEGImages
├── PairAnnotation
├──── trn 
├──── test
├──── val

Aachen (testing only)

  1. Download Aachen images dataset from here and pairs split from here. The structure should be:
Aachen
├── aachen_test_1500_pairs.txt
├── database_intrinsics.txt
└── db_poses.txt
└── images_upright

Megadepth (training and testing)

  1. Download Megadepth dataset following the instructions from LoFTER here. If you want to do testing only, skip this step.
  2. Download the test scene info from here.
  3. After extraction, the structure should be:
megadepth
├── train-data (for training)
└────phoenix
└────megadepth_indices
└──────scene_info_0.1_0.7
└────────0000_0.1_0.3.npz
└────────...
├── megadepth_test_1500 (for testing)
└──── 0015_0.1_0.3.npz
└──── 0015_0.3_0.5.npz
└──── 0022_0.1_0.3.npz
└──── 0022_0.5_0.7.npz
└──── Undistorted_SfM
└──── megadepth_test_1500.txt

Scannet (training and testing)

  1. Download scannet dataset from here. If you want to do testing only, skip this step.
  2. Download the test scene info from here.
  3. After extraction, the structure should be:
scannet
├── scannet_test_1500 (testing)
└────intrinsics.npz
└────scannet_test.txt
└────statistics.json
└────test.npz

TAPVID_DAVIS (testing only)

  1. Download TAPVID dataset from here. After extraction, no more action required. The structure should be:
tapvid_davis
├── tapvid_davis.pkl
├── README.md
└── SOURCES.md

COCO-20k (training only)

  1. Download the COCO-20k subset provided by XFeat from here. After extraction, no more action required. The structure should be:
coco_20k
├── 000000140092.jpg
└── ...jpg