Skip to content

Latest commit

 

History

History
129 lines (109 loc) · 7.02 KB

Format.md

File metadata and controls

129 lines (109 loc) · 7.02 KB

Format of the SIND dataset

                                          Fig.1 Ground Coordinate System and Traffic Lights Distribution
  1. Veh_smoothed_tracks.csv
    This file records the time-dependent vehicle trajectory and motion state parameters:
Column track_id frame_id timestamp_ms agent_type x y vx vy yaw_rad heading_rad length width ax ay v_lon v_lat a_lon a_lat
Unit [-] [-] [ms] [-] [m] [m] [m/s] [m/s] [rad] [rad] [m] [m] [m/s^2] [m/s^2] [m/s] [m/s] [m/s^2] [m/s^2]
  • track_id: the id number of the vehicle in this record
  • frame_id: the id number of the data frame in this record
  • timestamp_ms: The time corresponding to each data frame
  • agent_type: the specific class to which the vehicle belongs(car, truck, bus, motorcycle, bicycle, tricycle)
  • x : The x position of the center of the vehicle bounding box in the ground coordinate system
  • y : The y position of the center of the vehicle bounding box in the ground coordinate system
  • vx: The component of the center's velocity in the x-direction
  • vy: The component of the center's velocity in the y-direction
  • yaw_rad : the angle between the direction of the front longitudinal axis of the vehicle and the positive direction of the x-axis
  • heading_rad: the angle between the moving direction of the center of the vehicle bounding box and the positive x-axis
  • length: the length of the vehicle
  • width: the width of the vehicle
  • ax: The component of the acceleration at the center in the x direction
  • ay: The component of the acceleration at the center in the y direction
  • v_lon: Component of the center velocity in the direction of the vehicle's longitudinal axis
  • v_lat: Component of the center velocity in the direction of the vehicle's lateral axis
  • a_lon:Component of the center acceleration in the direction of the vehicle's longitudinal axis
  • a_lat:Component of the center acceleration in the lateral direction of the vehicle
  1. Ped_smoothed_tracks.csv
    Similar format to Veh_smoothed_tracks.csv, but pedestrians are treated as points without scale and orientation:
Column track_id frame_id timestamp_ms agent_type x y vx vy ax ay
Unit [-] [-] [ms] [-] [m] [m] [m/s] [m/s] [m/s^2] [m/s^2]
  • track_id:the id number of the pedestrain in this record P1,P2,...,Pn
  • agent_type: Only pedestrian
  • Other items have the same meaning as the items with the same name in the Veh_smoothed_tracks.csv file.
  1. Veh_tracks_meta.csv
    This file provides the metadata of vehicle track, mainly including vehicle attributes and track labels.
Column trackId initialFrame finalFrame Frame_nums length width class CrossType Signal_Violation_Behavior
Unit [-] [-] [-] [-] [m] [m] [-] [-] [-]
  • trackId: the id number of the vehicle in this record
  • initialFrame: The first data frame ID that the vehicle begins to appear in the record
  • finalFrame: The last data frame ID of the vehicle in the record
  • Frame_nums: The total number of data frames of vehicle track, which is equal to finalframe - initialframe + 1
  • length: the length of the vehicle
  • width: the width of the vehicle
  • class: One of cars, trucks, buses, motorcycles, bicycles and tricycles
  • CrossType: The class of vehicles passing through the intersection, which are classified as StraightCross, LeftTurn, RightTurn, and Others( not necessarily passing through the intersection, maybe some special action such as driving to the sidewalk)
  • Signal_Violation_Behavior: The three labels are red-light running, yellow-light running, No violation of traffic lights, the first two are only for vehicles with left turn and straight ahead behavior. (In China, right turning vehicles are allowed to run red lights)
  1. Ped_tracks_meta.csv
    The behavior of pedestrians is not easy to describe, so we do not provide labels for their behavior type and violation type.
Column trackId initialFrame finalFrame Frame_nums
Unit [-] [-] [-] [-]
  • trackId: the id number of the pedestrain in this record P1,P2,...,Pn
  • class: Only pedestrain
    Other items have the same meaning as the items with the same name in the Veh_tracks_meta.csv file
  1. TrafficLight_[record_name].csv
    This document records the states of traffic lights over time:
Column RawFrameID timestamp(ms) Traffic light X
Unit [-] [ms] [0,1,3]
  • FrameID: The original frame Id corresponding to the moment when the state of the traffic light changes(The raw frame corresponding to the data frame: Frame * 3 = RawFrameID)
  • timestamp（ms）: The moment when the state of the traffic light changes
  • Traffic light X(1-8): 0 for red light status, 1 for green light status, 3 for yellow light status; the traffic light position is as shown in the Fig.1.
  1. Recording_metas.csv
    This file records some information related to data collection.
Column RecordingID City Record weekday Record time period Weather Raw frame rate Record duration Tps_num number of categories
Unit [-] [-] [-] [-] [-] [Hz] [s] [-] [-]
  • RecordingID: the record's number among all records
  • City: The city where the data is located. Only Tianjin is in all records of SIND.
  • Record weekday: The weekday on which the data was collected.
  • Record time period: This record corresponds to a specific time period, accurate to the hour.
  • Weather:Sunny, cloudy, or after rain.
  • Raw frame rate: Fixed at 29.97hz
  • Record duration:Total duration of the record(Unit: second).
  • Tps_num: Total number of traffic participants in this record
  • number of categories(car, truck, bus, bicycle, motorcycle, tricycle, pedestrian):The number of traffic participants in each category in this record

In addition to the record file, we also provide a high-definition map of the intersection in Lanelet2 format(see Fig.2), the origin of the map coincides with the origin of the ground coordinate system.

                                                Fig.2 Semantic HD-map in lanelet2 format