Skip to content

This is the offical repository of INFRA-3DRC-Dataset, public dataset of 3D automotive Radar and RGB camera recorded using Intelligent roadside Infrastructure setup

Notifications You must be signed in to change notification settings

FraunhoferIVI/INFRA-3DRC-Dataset

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

52 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

INFRA-3DRC Dataset

This is the offical repository of INFRA-3DRC Dataset, a public dataset comparised of 3D automotive Radar and RGB camera data generated using Intelligent roadside Infrastructure (also known as smart Infrastructure) setup. It contains the links to download dataset, software development kit (SDK) to read and perform basic functions on the dataset, and related documentation.

This work is supported by the Bavarian Ministry of Economic Affairs, Regional Development and Energy (StMWi), Germany within the Project “InFra — Intelligent Infrastructure.”

Disclaimer: In order to comply with European GDPR data privacy, the RGB camera images are annonymized to hide human faces and vehicle number plates in the camera images where this information was clearly visible. In other frames, where the information is self annonymized due to distance from camera, no annonymization is applied. For this purpose, the state-of-the-art algorithms known best to the knowledge of the developer are used.

Overview


Introduction

INFRA-3DRC Dataset is the public dataset generated using smart infrastructure setup. It consists of calibrated, time-synchronized, and annotated sensor frames of 3D autotmotive radar and RGB mono-camera. Each radar frame contains point-wise annotation and each camera image is annotated in form of 2D bounding box(es). For completeness, calibrated and time-synchronized 3D lidar sensor frames are also included in the dataset, but these frames are not annotated.

Dataset consists of 25 scenes recorded from different geographical locations in different light conditions - day light, twilight and night. It contains total 2768 frames of each camera, radar, and lidar sensors. Camera annotations have 4172 instances and radar annotations have 4074 instances of total six object categories - adult, group, bicycle, motorcycle, car, and bus. It contains 22966 labeled points of radar data of valid traffic road users.

Some examples of the dataset are shown below.

twilight scene

day time pedestrain crossing junction scene

night time scene

open parking scene


Smart Infrastructure sensors setup

  • The setup consists of

    • RGB mono camera (ids CP-5260 rev 2 with 8 mm lens C-mount)
    • 3D Automotive radar (continental LRR ARS548)
    • 3D-Lidar (Ouster OS1-64 channels),
  • All the sensors are mounted on the extendable tripod for modular and flexible data collection.

  • Below images are subject to copyright.


SDK User Guide

SDK as python package can be downloaded from infra-3drc on pypi.
The associated source code is hosted at GITHUB-REPO.
SDK user guide with examples is available at SDK-USER-GUIDE.


Dataset Format

  ├── INFRA-3DRC_scene-01
    ├── camera_01
      ├── camera_01__annotation - contains json files for each camera image
      ├── camera_01__data - contains png files of camera images
    ├── radar_01
      ├── radar_01__annotation - contains json files for each radar frame
      ├── radar_01__data - contains pcd files for each radar frame
    ├── lidar_01
      ├── lidar_01__data - contains pcd files for each lidar frame
    ├── calibration.json - contains extrinsic and intrinsic calibration matrix
    └── scene.json - contains meta-data of the scene

  ├── INFRA-3DRC_scene-02
    ├── camera_01
      ├── camera_01__annotation
      ├── camera_01__data
    ├── radar_01
      ├── radar_01__annotation
      ├── radar_01__data
    ├── lidar_01
      ├── lidar_01__data
    ├── calibration.json
    └── scene.json
.....

  ├── INFRA-3DRC_scene-NN
    ├── camera_01
      ├── camera_01__annotation
      ├── camera_01__data
    ├── radar_01
      ├── radar_01__annotation
      ├── radar_01__data
    ├── lidar_01
      ├── lidar_01__data
    ├── calibration.json
    └── scene.json

Dataset

To download the dataset, refer the following page DATASET-DOWNLOAD


Annotation Format

Dataset contains annotations of 3D automotive radar sensor and RGB camera images. For details refer the following page ANNOTATION-DETAILS


License

  • The dataset and the associated code is realeased under the CC BY-NC 4.0. See here
  • Under this license, it is allowed to download and use only for research, teaching, and academic purpose only.
  • Use of this dataset and associated code is prohibited to use for any commercial application.
  • If in case, it is needed for commercial use, then kindly contact us (see bottom of this page).

Citation

@ARTICLE{10459049,
  author={Agrawal, Shiva and Bhanderi, Savankumar and Elger, Gordon},
  journal={IEEE Access}, 
  title={Semi-Automatic Annotation of 3D Radar and Camera for Smart Infrastructure-Based Perception}, 
  year={2024},
  volume={12},
  pages={34325-34341},
  doi={10.1109/ACCESS.2024.3373310}}

@misc{infra_3drc,
  author = {Shiva Agrawal and  Savankumar Bhanderi},
  title = {INFRA-3DRC dataset},
  version = {1.0.0},
  year = {2024},
  url = {https://github.com/FraunhoferIVI/INFRA-3DRC-Dataset},
  orcid = {0000-0001-8633-341X, 0000-0001-7257-6736},
  doi = {10.24406/fordatis/297},
  license = {CC BY-NC 4.0}
}

Contact Us

Shiva Agrawal, [email protected]
Savankumar Bhanderi, [email protected]

About

This is the offical repository of INFRA-3DRC-Dataset, public dataset of 3D automotive Radar and RGB camera recorded using Intelligent roadside Infrastructure setup

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages