SLAM-generation

Generate simple data for SLAM estimators.

Data sources

We used groundtruth data from snapdragon dataset #3 (groundtruth.txt), which have to be placed in data folder.

Camera calibration

both camera calibrations (mono and stereo) use same intrinsic:
- fx, fy = 420, 420
- skew = 0
- u0, v0 = 320, 240
baseline (stereo) = 0.2

Data files

files stored in data folder used for generation

Groundtruth

groundtruth.txt contains groundtruth data from snapdragon dataset.
file have one header line
data are decimal numbers separated with space:
- timestamp in seconds
- three numbers representing coordinates
- four numbers representing quaternion (x, y, z, w)

Walls

walls.csv
contains information used for generating point clouds
data are decimal numbers separated with space:
- three numbers representing coordinates of the corner of the wall
- three numbers representing coordinates of the first adjenct corner
- three numbers representing coordinates of the second adjenct corner
- number of points in the given wall (numbers' decimal part will be removed)

Output positions

output_positions.csv contains the ground truth positions
data are three decimal numbers representing coordinates separated with space

Output points

output_points.csv contains the positions of the points
data are three decimal numbers representing coordinates separated with space

Output seen

output_seen_{mono|stereo}.csv contains visible positions
data are two integers separated with space:

first integer represents the line number of the position used for testing, whether the camera sees the point (from output_positions.csv)
second integer represents the line number of the point that was seen by the camera (from output_points.csv)

Output projections

output_projections_{mono|stereo}.csv contains all generated measurements
data are four decimal numbers:

first is timestamp in second based on groundtruth time used for simulation
last values are from the projections in gtsam (x, y) | (uL, uR, v)

Scripts

. scripts/run.sh generate data