## Sensor Fusion: Camera, Lidar and Bus Dataset

The Sensor Fusion dataset contains three recorded sequences in three
locations in Germany:

Gaimersheim
Ingolstadt
Munich

Each sequence has readings from the camera, lidar, and bus data of the
vehicle. In total there are 392,556 frames. This data is not annotated.

Each frame contains the following items:

- RGB image
- registration info file
- 3D point cloud

All frames are grouped in three different scenes with each scene contained
in its corresponding folder. Scene folder names are in the
'YYYYMMDD_hhmmss' format. They represents the date and time of the
recording. Each scene is further divided into two folders:

- 'camera': input images and json info files
- 'lidar': input 3D point clouds

Each of these folders are further divided depending on the camera from
which the data was recorded. There are six cameras available in the
vehicle, therefore, the following are the camera folders:

- 'cam_front_center'
- 'cam_front_left'
- 'cam_front_right'
- 'cam_side_left'
- 'cam_side_right'
- 'cam_rear_center'

Lastly, each of these folders contains the corresponding item for each
frame.

These are the filename formats for each item of a single sample:

- input RGB image  : YYYYMMDDhhmmss_camera_[frontcenter|frontleft|frontright|sideleft|sideright|rearcenter]_[ID].png
- input info       : YYYYMMDDhhmmss_camera_[frontcenter|frontleft|frontright|sideleft|sideright|rearcenter]_[ID].json
- 3D lidar pcloud  : YYYYMMDDhhmmss_lidar_[frontcenter|frontleft|frontright|sideleft|sideright|rearcenter]_[ID].npz

For example a sample with ID 1617 from a scene recorded on 2018-08-07
14:50:28 from the front center camera would be have the following items
in these locations:

- input RGB image  : 20180807_145028/camera/cam_front_center/20180807145028_camera_frontcenter_000001617.png
- input info       : 20180807_145028/camera/cam_front_center/20180807145028_camera_frontcenter_000001617.json
- 3D lidar pcloud  : 20180807_145028/lidar/cam_front_center/20180807145028_lidar_frontcenter_000001617.npz

For further explanations regarding the format of each of these items,
please refer to the tutorial in our dataset web page.

# Bus Data

The vehicle bus singals are all stored in a single json file for all the frames.
The bus signal file can be found in:

YYYYMMDD_hhmmss/bus/YYYYMMDDhhmmss_bus_signals.json

The following signals are recorded:

- acceleration_x
- acceleration_y
- acceleration_z
- accelerator_pedal
- accelerator_pedal_gradient_sign
- angular_velocity_omega_x
- angular_velocity_omega_y
- angular_velocity_omega_z
- brake_pressure
- distance_pulse_front_left
- distance_pulse_front_right
- distance_pulse_rear_left
- distance_pulse_rear_right
- latitude_degree
- latitude_direction
- longitude_degree
- longitude_direction
- pitch_angle
- roll_angle
- steering_angle_calculated
- steering_angle_calculated_sign
- vehicle_speed

Each sample is associated with a timestamp and each signal has its unit
which is specified in the json file.