V2x SIM dataset (Shanghai Jiaotong University & New York University)

V2X-Sim summary

One 、 brief introduction

V2X-Sim,Vehicle-to-Everything Short for simulation , By New York University AI4CE Laboratory and Shanghai Jiaotong University MediaBrain The first synthetic developed by the team V2X Assisted autopilot collaborative sensing data set , It aims to stimulate the research of multi-agent multi-modal multi task perception . because V2X The immaturity of and the cost of operating multiple autonomous vehicle at the same time , Building such a real-world data set for the research community is very expensive and laborious . therefore , Use highly realistic CARLA-SUMO Joint simulation to ensure that our data set is representative compared with real-world driving scenes .

Two 、 Sensor setup and data collection

1. sensor

（1） Each vehicle is equipped with 6 individual RGB / depth / Semantic segmentation camera 、1 individual BEV Semantic segmentation camera 、1 individual LiDAR and semantics LiDAR sensor 、1 individual IMU and 1 individual GNSS sensor .

（2） by the road , We will 4 individual RGB camera 、 One BEV Semantic segmentation camera 、 One LiDAR And a semantics LiDAR Connect to infrastructure . Cameras on vehicles and roadside infrastructure can cover 360°, To ensure all-round perception .

2. Map

The data is centralized CARLA Of Town 3、Town 4 and Town 5. Use the above CARLA-SUMO Joint simulation to generate multiple types of vehicles , Like cars 、 Bicycle 、 Motorcycles and emergency vehicles , To retrieve sensor data from the record file .

3、 ... and 、 Use of data sets

1. precondition ： Need to download nuScenes devkit package

2. Load data set

from  nuscenes.nuscenes  import  NuScenes 
import  matplotlib.pyplot  as  plt 
import  numpy  as  np 
import  os 
from  IPython.display  import  Image 
import  collections

in_path = 'V2X-Sim Dataset path '

nusc = NuScenes(version='v1.0-mini', dataroot=in_path, verbose=True)

3. Scene information

# To investigate a single scene

my_scene = nusc.scene[0]
my_scene

4.sample Information

my_scene = nusc.scene[0]
first_sample_token = my_scene['first_sample_token']
my_sample = nusc.get('sample', first_sample_token)

my_sample

5. Visualization of sensor data

Because the sensor data set is too large to download , All without practice ！

Four 、 appendix

Reference from ：https://ai4ce.github.io/V2X-Sim/