Skip to content

CarlaScapes dataset: Vehicle semantic segmentation. CarlaScapes dataset samples from CARLA simulator involving 12 different Map. Specifically, these maps imitate the real-world settings such as urban, suburban, highway, and rural areas.

Notifications You must be signed in to change notification settings

yjzhai-cs/CarlaScapes

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CarlaScapes dataset: Vehicle semantic segmentation

Dataset Overview

CarlaScapes dataset samples from CARLA simulator involving 12 different Maps(Town01-Town07, Town10HD, Town11-Town13, Town15). Specifically, these maps imitate the real-world settings such as urban, suburban, highway, and rural areas. The dataset contains 5000 images with 19 semantic categories and 9 custom categories. The dataset is divided into 4000 training images and 1000 testing images. In the following, we give an example of different map's images.

Features

The semantic categories are divided into 19 cityscapes classes (bold type) and 9 custom classes (italic form). The 19 cityscapes classes are the same as in the CityScapes dataset.

Tag Color LabelId Tag Color LabelId Tag Color LabelId
unlabeled (0,0,0) 0 terrain (152,251,152) 10 static (110,190,160) 20
road (128,64,128) 1 sky (70,130,180) 11 dynamic (170,120,50) 21
sidewalk (244,35,232) 2 pedestrian (220,20,60) 12 other (55,90,80) 22
building (70,70,70) 3 rider (255,0,0) 13 water (45,60,150) 23
wall (102,102,156) 4 car (0,0,142) 14 road line (157,234,50) 24
fence (190,153,153) 5 truck (0,0,70) 15 ground (81,0,81) 25
pole (153,153,153) 6 bus (0,60,100) 16 bridge (150,100,100) 26
traffic_light (250,170,30) 7 train (0,80,100) 17 rail track (230,150,140) 27
traffic_sign (220,220,0) 8 motorcycle (0,0,230) 18 guard rail (180,165,180) 28
vegetation (107,142,35) 9 bicycle (119,11,32) 19

Download

We also provide two data compression files for you to choose from:

After downloading and unzipping the above files, you can get the following directory structure. Concretely, the file naming format is {town}_{timestamp}_{frame}_{type}.{suffix}.

CarlaScapes
├── train
│   ├── Town01
│   │   ├── Town01_000010_074800_bounding_box.xml
│   │   ├── Town01_000010_074800_color.png
│   │   ├── Town01_000010_074800_gnss.json
│   │   ├── Town01_000010_074800_img.png
│   │   ├── Town01_000010_074800_instance.json
│   │   `── Town01_000010_074800_labelIds.png
│   ├── Town02
│   │   ├── ...
├── test
│   ├── Town15
│   │   ├── ...
`── README.md

Reference

[1] Cityscapes Dataset

[2] CARLA Simulator

[3] carla-semantic-localization

About

CarlaScapes dataset: Vehicle semantic segmentation. CarlaScapes dataset samples from CARLA simulator involving 12 different Map. Specifically, these maps imitate the real-world settings such as urban, suburban, highway, and rural areas.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published