LiCAR dataset (v1) - LiDAR CAR dataset

This dataset contains Spherical Range Images (SRI) obtained from a LiDAR sensor. It is composed of pseudo-RGB images in several environments and it can be used to train and evaluate car detection and segmentation methods. Each image contains on average 14 instances in different conditions, including variance of depth, illumination and occlusion. There are 400 images with 5652 instances labeled according to the car class.

The next table shows the sample distribution for each class:

Class #samples Description
Car 400 Car category.

For evaluation we have provided the dataset separated in training, validation and testing sets, following the 85/10/5 proportion, resulting in the following distribution:

Class #samples train set #samples validation set #samples test set
Car 340 40 20

All environments are in the University of Alicante area and include different backgrounds.

1 2
1 2

LICENSE AND ACCESS

This dataset is shared only for non-profit research or educational purposes. If you use this dataset or a part of it, please respect these terms of use and reference the original work in which it was published.

FILE FORMAT

The images were acquired with a Ouster OS1-128 3D-LiDAR sensor. From them, 3 channels were combined in order to generated the pseudo-RGB images. These final images have 2048 x 128 pixels. The images are stored as PNG where pixel values represent fake RGB colors. The distance between cars and camera has been changed in order to obtain closer and further samples.

All the images are distributed in train / validation / test sets following the aforementioned 85 / 10 / 5 distribution.

In addition, the bounding box and segmentation area of each object is included. This labeling is stored in a JSON file, one for each image. For being able to convert this dataset to COCO format, we need to use the tool labelme2coco.py, which is inside the labelme repository.

RELATED PUBLICATIONS (CITATION)

Please, if you use this dataset or part of it, cite the following publication: 

  @unpublished{VelascoND,
    author = {E.P Velasco-Sánchez, I-L Páez-Ubieta and S.T. Puente},
    title = { LiCAR: pseudo-RGB LiDAR image for CAR segmentation },
    note = {Submitted to 5th International Conference on Robotics, Computer Vision and Intelligent Systems (ROBOVIS 2025)},
    year = {N.D.},
  }

Research work was funded by grant PID2021-122685OB-I00 funded by MICIU/AEI/10.13039/501100011033 and ERDF/EU and grand PRE2019-088069 funded by MICIU/AEI/10.13039/501100011033 and ESF Investing in your future. The computer facilities were provided through the IDIFEFER/2020/003 project.

DOWNLOAD

To download this dataset fill in the following form.

VERSIONS

V1 - Released in October 2024. This version of the dataset contained 400 images separated in train, validation and test sets. The dataset has a JSON file for each image.