Indoor3Dmapping dataset

Data Organization
Under the root directory for the whole acquisition, there is a positions.csv file and 3 subdirectories: img, dense, and sparse. The mobile mapping 3D dataset was generated walking around an indoor space and each <positionID> corresponds to a unique pose along the trajectory of this motion. This version of the dataset contains a total of 99 unique poses. There is a separation of 1 meter between each adjacent pose.

root
├── img
│  ├── <positionID>.jpg
|  └── ...
├── dense
│  ├── <positionID>.png
|  └── ...
├── sparse
│  ├── <positionID>.png
|  └── ...
└── positions.csv

positions.csv

• File format: One ASCII file.
• File structure Rows: Each image is one record.
• File structure Columns: Comma separated headers, with exact order described below.
  • Filename, column 0: Panorama file name as on disk, without file extension.
  • Timestamps, column 1: Absolute time at which the panorama was captured, Decimal notation, without thousands separator (microseconds).
  • X,Y,Z, columns 2 through 4: Position of the panoramic camera in decimal notation, without thousands separator (meters).
  • w,x,y,z, columns 5 through 8: Rotation of the camera, quaternion.

sparse

• Set of equirectangular rendered depth images.
• 1920x960 resolution
• 16-bit grayscale PNG
• White → 0 m
• Black → ≥ 16 m or absent geometry
• Occlusions: If a pixel was hit by several rays, only the value of the closest one is represented. 

dense

• Set of equirectangular rendered depth images.
• 1920x960 resolution
• 16-bit grayscale PNG
• White → 0 m
• Black → ≥ 16 m or absent geometry
• Occlusions: If a pixel was hit by several rays, only the value of the closest one is represented.

img
A set of equirectangular panoramic images was taken with a 360° color camera in 1920x960 resolution. They follow the same trajectory.