Download ScaNnet200 and follow Mask3D to preprocess data. A preprocessed point cloud scene has the following format id.npy. Kindly move all preprocessed scenes to their corresponding folders e.g. 0011_00.npy to ./data/scannet200/scene0011_00.
ScanNet200 data folder should have the following structure
./data
└── scannet200
└── ground_truth
├── scene0011_00.txt
└── ...
└── scene0011_00
├── poses <- folder with camera poses
│ ├── 0.txt
│ ├── 1.txt
│ └── ...
├── color <- folder with RGB images
│ ├── 0.jpg (or .png/.jpeg)
│ ├── 1.jpg (or .png/.jpeg)
│ └── ...
├── depth <- folder with depth images
│ ├── 0.png (or .jpg/.jpeg)
│ ├── 1.png (or .jpg/.jpeg)
│ └── ...
├── intrinsics.txt <- camera intrinsics
│
└── 0011_00.npy <- preprocessed point cloud of the scene
│
└── scene0011_00_vh_clean_2.ply <- raw point cloud of the scene
└── ...
Run the following command.
sh scripts/get_replica_dataset.sh
A replica folder with the following format will be generated.
./data
└── replica
└── ground_truth
├── office0.txt
└── ...
└── office0
├── poses <- folder with camera poses
│ ├── 0.txt
│ ├── 1.txt
│ └── ...
├── color <- folder with RGB images
│ ├── 0.jpg (or .png/.jpeg)
│ ├── 1.jpg (or .png/.jpeg)
│ └── ...
├── depth <- folder with depth images
│ ├── 0.png (or .jpg/.jpeg)
│ ├── 1.png (or .jpg/.jpeg)
│ └── ...
├── intrinsics.txt <- camera intrinsics
│
└── office0_mesh.ply <- raw point cloud of the scene
└── ...