Argoverse2 数据集分析

Argoverse2 数据集分析

import pandas as pd
from av2.map.map_api import ArgoverseStaticMap
from pathlib import Path
id='f9c5274f-2487-41cc-a8c7-5294325db629'
raw=f'/mnt/ve_share2/zy/Argoverse_2_Motion_Forecasting_Dataset/raw/val/{id}/scenario_{id}.parquet'
map_path=f'/mnt/ve_share2/zy/Argoverse_2_Motion_Forecasting_Dataset/raw/val/{id}/log_map_archive_{id}.json'
static_map_path =Path(map_path)
df = pd.read_parquet(raw)
static_map = ArgoverseStaticMap.from_json(static_map_path)

scenario_{id}.parquet：

​
observed:bool,是否在观测区间

track_id:每个交通参与者特有的id,自车是AV

object_type:对应每个交通参与者的类型，如vehicle

object_category:给轨迹序列分配种类，用于给轨迹预测的数据质量提供参考，一般来说，有四种：SCORED_TRACK，UNSCORED_TRACK，FOCAL_TRACK，TRACK_FRAGMENT。其中FOCAL_TRACK和SCORED_TRACK数据质量较好，UNSCORED_TRACK用于当作上下文输入，数据质量一般，而TRACK_FRAGMENT的时间长度不定，数据质量较差

TRACK_FRAGMENT: Lower quality track that may only contain a few timestamps of observations. 在数据中以整数0表示
UNSCORED_TRACK: Unscored track used for contextual input. 在数据中以整数1表示
SCORED_TRACK: High-quality tracks relevant to the AV - scored in the multi-agent prediction challenge. 在数据中以整数2表示
FOCAL_TRACK: The primary track of interest in a given scenario - scored in the single-agent prediction challenge. 在数据中以整数3表示

timestep:时间步 [0:110)

position_x,position_y:物体bounding box的中心坐标

heading:地图坐标系下的物体bounding box的航向角，单位是弧度

velocity_x,velocity_y: 物体的速度

scenario_id:每个场景特有的id

start_timestamp,end_timestamp:开始，结束的时间

num_timestamps:所有的时间戳 110

focal_track_id:焦点agent(focal agent)的track ID

city：对应的城市

log_map_archive_{id}.json

ArgoverseStaticMap(log_id=‘f9c5274f-2487-41cc-a8c7-5294325db629’, ##场景对应的id

vector_drivable_areas= #行驶区域

{

24214204: DrivableArea(id=24214204, area_boundary=[Point(x=5778.46, y=-295.23, z=-23.72), Point(x=5773.76, y=-287.93, z=-23.71)]), ## area_boundary 3D多边形，用来表示可行驶区域的边界

24214207: DrivableArea(id=24214207, area_boundary=[Point(x=5849.63, y=-328.19, z=-24.09), Point(x=5847.07, y=-330.65, z=-24.04)]),

24214212: DrivableArea(id=24214212, area_boundary=[Point(x=5887.84, y=-448.97, z=-23.16), Point(x=5888.75, y=-450.0, z=-23.14)])

},

vector_lane_segments=

{

418721305: LaneSegment(id=418721305, is_intersection=False, lane_type=<LaneType.BIKE: ‘BIKE’>, right_lane_boundary=Polyline(waypoints=[Point(x=5899.93, y=-450.0, z=-23.55), Point(x=5869.51, y=-401.51, z=-23.41)]), left_lane_boundary=Polyline(waypoints=[Point(x=5898.41, y=-450.0, z=-23.48), Point(x=5868.5, y=-402.24, z=-23.35)]), right_mark_type=<LaneMarkType.NONE: ‘NONE’>, left_mark_type=<LaneMarkType.SOLID_WHITE: ‘SOLID_WHITE’>, predecessors=[418721433], successors=[418721716], right_neighbor_id=None, left_neighbor_id=418721775),

418721392: LaneSegment(id=418721392, is_intersection=False, lane_type=<LaneType.VEHICLE: ‘VEHICLE’>, right_lane_boundary=Polyline(waypoints=[Point(x=5865.73, y=-404.23, z=-23.19), Point(x=5859.68, y=-394.59, z=-23.19)]), left_lane_boundary=Polyline(waypoints=[Point(x=5862.82, y=-406.0, z=-23.02), Point(x=5856.71, y=-396.45, z=-22.98)]), right_mark_type=<LaneMarkType.DASHED_WHITE: ‘DASHED_WHITE’>, left_mark_type=<LaneMarkType.NONE: ‘NONE’>, predecessors=[418722058], successors=[418721569], right_neighbor_id=418721834, left_neighbor_id=None),

},

is_intersection:表示该lane segment是否位于一个路口内

lane_type：表示车道线类型

right_lane_boundary，left_lane_boundary：3D线条，表示右，左车道边界线

right_mark_type， left_mark_type：右，左车道边界线的线型，包括：NONE,SOLID_WHITE,DASHED_WHITE，SOLID_YELLOW,…

predecessors:该lane segment的前继lane segment的unique ID

right_neighbor_id:该lane segment的右邻lane segment的unique ID

left_neighbor_id:该lane segment的左邻lane segment的unique ID

vector_pedestrian_crossings=

{

25571845: PedestrianCrossing(id=25571845, edge1=Polyline(waypoints=[Point(x=5842.09, y=-356.41, z=-23.48), Point(x=5823.69, y=-368.34, z=-23.18)]), edge2=Polyline(waypoints=[Point(x=5840.76, y=-352.91, z=-23.56), Point(x=5821.85, y=-365.18, z=-23.16)])),

25572551: PedestrianCrossing(id=25572551, edge1=Polyline(waypoints=[Point(x=5850.71, y=-340.33, z=-23.8), Point(x=5862.61, y=-340.18, z=-23.92)]), edge2=Polyline(waypoints=[Point(x=5852.44, y=-336.47, z=-23.89), Point(x=5860.4, y=-336.32, z=-23.94)])),

25572666: PedestrianCrossing(id=25572666, edge1=Polyline(waypoints=[Point(x=5837.81, y=-341.13, z=-23.78), Point(x=5843.15, y=-349.68, z=-23.71)]), edge2=Polyline(waypoints=[Point(x=5833.17, y=-340.49, z=-23.66), Point(x=5840.82, y=-352.41, z=-23.57)]))

},

edge1,edge2:3D多边形，代表人行道的其中一边，由2个waypoints组成

raster_drivable_area_layer=None, raster_roi_layer=None, raster_ground_height_layer=None)

每个scenario的局部地图包含距离ego-vehicle的轨迹100m l2距离内的所有实体