Visual-Acoustic Vehicle Dataset

On the Development of an Acoustic-Driven Method to Improve Driver's Comfort Based on Deep Reinforcement Learning

Access our publication: Link

The Visual-Acoustic Vehicle Dataset is a set of multimodal data collected from a Lincoln MKS car including vehicle sensors information, 360º image data collected from eight cameras on top of the car, and LiDAR information. All the data has timestamps for syncronization purposes and is divided in four experiments.

Info	Description	Sampling Rate

nmea_sentence	GPS sentences with the data type GGA	182Hz
vehicle/acc_ped_eng	Includes throttle_rate, throttle_pc, and engine_rpm (revolutions per minute)	101Hz
vehicle/brake_ped	Shows if brake pedal is being pressed	50Hz
vehicle/brake_torq	Includes brake_torque_request, brake_torque_actual, and vehicle_speed	50Hz
vehicle/gear	Which gear the vehicle is currently using	50Hz
vehicle/imu/data_raw	Angular_velocity, and linear_acceleration along with the covariance	50Hz
vehicle/joint_states	Position and velocity of each wheel and the steer on the front left and front right	116Hz
vehicle/steering_ang	Steering wheel angle	50Hz
vehicle/steering_torq	Steering wheel torque	50Hz
vehicle/suspension	Front and rear suspension	50Hz
vehicle/tire_press	Tire pressure for each wheel	2Hz
vehicle/turn_sig	Whether turn signal was being used or not	1Hz
vehicle/twist	Linear and angular twist of entire vehicle	50Hz
vehicle/wheel_speeds	Records the speed of each of the four wheels in miles per hour	100Hz

Vehicle data

Tables containing car info topics as steering wheel angle, brake pedal status, tire pressure for each wheel, and IMU. (See all topics and sampling rates here)

Camera and Audio data

Images from the eight Logitech C922 video cameras pointing outwards towards the street in a circular formation to get a 360º view from the vehicle. Audio data from the Olympus ME-51S microphones placed outside and inside the vehicle.

Audio only: 1.5GB (Images coming soon)

LiDAR data

Depth information of the car surroundings collected from a Velodyne 64E-S2.

Coming Soon