LiDAR-CS

LiDAR Dataset with Cross-Sensors (LiDAR-CS Dataset), which contains large-scale annotated LiDAR point cloud under 6 groups of different sensors but with the same correspondence scenarios, captured from hybrid realistic LiDAR simulator. As far as we know, LiDAR-CS Dataset is the first dataset focused on the sensor (e.g., the points distribution) domain gaps for 3D object detection in real traffic.

Update:

We also provide point cloud data with different sensor heights (1.0m, 1.5m, 2.0m, 2.5m, 3.0m).

Download

This is the official GitHub repository for LiDAR-CS dataset.

1. The download links can be found in download_data.sh
2. We also support Baidu Yunpan. Link：https://pan.baidu.com/s/1NyPziUeqfBSv6rgZBUspkQ with extraction code: x6o2

Data Sample

Getting Started

1. Data Prepare

Uncompress all the compressed files, for example

# For the normal compressed file
tar zxf VLD-16.tar.gz

# For the compressed file which is split due to the size limitation
cat VLD-128.tar.gz* | tar -zxf -

All the file will be organized as,

├── Livox
│   ├── bin
│   └── label
├── ONCE-40
│   ├── bin
│   └── label
├── VLD-128
│   ├── bin
│   └── label
├── VLD-16
│   ├── bin
│   └── label
├── VLD-32
│   ├── bin
│   └── label
└── VLD-64
│   ├── bin
│   └── label
└── splits
    ├── test.txt
    └── train.txt

2. Load Data

We follow KITTI to store the point cloud into binary files and the annotation results are stored in text files that are easy to parse.

Here is a python sample code to load the point cloud and the annotation file.

import numpy as np

def get_label(label_file):
    labels = np.loadtxt(label_file, dtype=str)
    if len(labels.shape) == 1:
        labels = labels[None, :]
    types, labels = labels[:, 0], labels[:, 1:].astype(np.float32)
    return types, labels

pc_path = 'VLD-16/bin/000000.bin'
label_path = 'VLD-16/label/000000.txt'

# the (x, y, z, intensity) are stored in binary
xyzi = np.fromfile(pc_path, dtype=np.float32).reshape(-1, 4) 

# types store the class names for the objects
# labels store a n * 7 ndarry and 7 is for (x, y, z, lenght, width, height, angle) in LiDAR coordinate.
types, labels = get_label(label_path) 

Todo List

• More sensors will be supported.
• Update the evaluation code.
• Update the cross evaluation code.

Changelog

• v1.0: support 6 sensors, including VLD-16/32/64/128, Livox and ONCE-40
• v1.1: support different sensor heights (1.0m, 1.5m, 2.0m, 2.5m, 3.0m) for VLD-64

Citation

If you find this dataset useful in your research, please consider cite:

@article{fang2023lidar,
  title={LiDAR-CS Dataset: LiDAR Point Cloud Dataset with Cross-Sensors for 3D Object Detection},
  author={Fang, Jin and Zhou, Dingfu and Zhao, Jingjing and Tang, Chulin and Xu, Cheng-Zhong and Zhang, Liangjun},
  journal={arXiv preprint arXiv:2301.12515},
  year={2023}
}