Pointillism: A Multi-modal dataset for automotive radar sensing.

Official multi-radar Dataset release for Pointillism: Accurate 3D Bounding Box Estimation with Multi-Radars. Official code for the paper is available at RP-net.

Sensor setup

• 1 x 16-channel OS1 Ouster LiDAR
• 1 x RealSense D415 Camera (for RGB only)
• 2 x IWR1443BOOST 3T4R Radars

Dataset

About 300 frames are provided from each sensor for 48 different scenes. Each frame is time synchronized among all the sensors using system timestamps.

Data Visualizer SDK

Requirements for Visualizer

1. Download the dataset from https://drive.google.com/file/d/1C-Ryh5W5FLPenNgPUDNcNwiac-NnwU1A/view?usp=sharing in data folder.
2. Dataset in pointillism-multi-radar-data should follow this directory structure.

   pointillism-multi-radar-data
   └───data
       └───scene{#}       'Where {} is omitted and # is the folder number'
           └───lidar
               └───*.pcd
           └───radar_0
               └───*.csv
           └───radar_1
               └───*.csv
           └───images
               └───*.jpg
           └───label
               └───*.json
   

3. Download ffmpeg https://ffmpeg.org/download.html and add ffmpeg.exe to path environment variable (only required to create video).
4. Download open3d, http://www.open3d.org/download/, for 3D visualization.

To run Visualizer:

To see all options type 
  python visualization.py -h

usage: visualization.py [-h] 
                        --type  [{Lidar3D,LidarBird,Camera,Radar3D,RadarBird}]
                        [--radar [RADAR [RADAR ...]]] 
                        [--video] 
                        --frame [FRAME [FRAME ...]] 
                        --dataset [DATASET [DATASET ...]]

Enter which file and type of image to be converted

optional arguments:
  -h, --help            show this help message and exit
  --type [{Lidar3D,LidarBird,Camera,Radar3D,RadarBird}]
                        --type LidarBird/Lidar3D/Camera/Radar3D/RadarBird
  --radar [RADAR [RADAR ...]]
                        --radar 1/0
  --video               Will create video of flag raised
  --frame [FRAME [FRAME ...]]
                        --frame int int
  --dataset [DATASET [DATASET ...]]
                        --dataset int

Example

python visualization.py --type LidarBird --frame 15 90 --dataset 16 --video

Citation

@inproceedings{bansal2020pointillism,
  title={Pointillism: accurate 3D bounding box estimation with multi-radars},
  author={Bansal, Kshitiz and Rungta, Keshav and Zhu, Siyuan and Bharadia, Dinesh},
  booktitle={Proceedings of the 18th Conference on Embedded Networked Sensor Systems},
  pages={340--353},
  year={2020}
}