ОбзорДокументацияВойти
/
VitalyyBezuglyj
/
ITLCampus-SM
Код
Запросы
0
Пакеты
0
Релизы
0
Аналитика
Безопасность

ITLCampus-SM

0

Описание

The dataset was recorded on the Husky robotics platform on the university campus and consists of 5 tracks recorded at different times of day (day/dusk/night) and different seasons (winter/spring).

5.16 MiB
README
License
В избранном0
Форки0
1
5
0
cfg
3 года назад
img
3 года назад
.gitignore
3 года назад
LICENSE
3 года назад
README.md
3 года назад
README.md

ITLCampus-SM

The dataset was recorded on the Husky robotics platform on the university campus and consists of 5 tracks recorded at different times of day (day/dusk/night) and different seasons (winter/spring).

Data

TrackSeasonTime of dayFrames, pcsFront cam, resBack cam, resLiDAR, rays6 DoF poseSemantic masks
00_2023-02-21winterday6206201920×10801920\times 10801920×10801920\times 108016☑front + back
1920×1080×651920\times 1080 \times 65 classes
01_2023-03-15winternight6266261920×10801920\times 10801920×10801920\times 108016☑front + back
1920×1080×651920\times 1080 \times 65 classes
02_2023-02-10wintertwilight6096091920×10801920\times 10801920×10801920\times 108016☑front + back
1920×1080×651920\times 1080 \times 65 classes
03_2023-04-11springday6386381920×10801920\times 10801920×10801920\times 108016☑front + back
1920×1080×651920\times 1080 \times 65 classes
11_2023-04-13springnight6316311920×10801920\times 10801920×10801920\times 108016☑front + back
1920×1080×651920\times 1080 \times 65 classes

6 DoF poses obtained using ALeGO-LOAM localization method refined with Interactive SLAM.

Sensors

SensorModelResolution
Front camZED (stereo)1920×10801920\times 1080
Back camRealSense D4351920×10801920\times 1080
LiDARVLP-1616×182416\times 1824

Semantics

Semantic masks are obtained using the Oneformer pre-trained on the Mapillary dataset.

The masks are stored as mono-channel images.Each pixel stores a semantic label. Examples of semantic information are shown in the table below:

LabelSemantic classColor, [r, g, b]
.........
10Parking[250, 170, 160]
11Pedestrin Area[96, 96, 96]
12Rail Track[230, 150, 140]
13Road[128, 64, 128]
.........

The complete list of semantic labels and their colors are described in the file anno_config.json.

An example of a mask over the image:

Structure

The data are organized by tracks, the length of one track is about 3 km, each track includes about 600 frames. The distance between adjacent frames is ~5 m.

The structure of track data storage is as follows:

where

  • ####
     - file name, which is the timestamp of the image/scan (virtual timestamp of the moment when the image/scan was taken)
  • .bin
     - files - LiDAR scans in binary format
  • .png
     - images and semantic masks
  • .csv
     - timestamp mapping for all data and 6DoF robot poses

An example of a track trajectory (track_map.png):