LiDAR Datasets point clouds visualization

This repo is a visualization implementation for KITTI dataset, Waymo open dataset and nuScense dataset in python.

It is tested with python 3.9.

Preparation

• Environments

pip install -r .\requirements.txt

To use nuScenes dataset

pip install nuscenes-devkit==1.1.9

To use waymo open dataset

pip install tensorflow==2.5.0 -i https://pypi.org/simple/

Custom Data

You can easily visualize your point clouds and bounding box by this repo with Nx3 (Nx4 if with intensity) np.array for point clouds and Nx7 np.array for 3d bounding boxes. There is a demo that can be a reference

python demo.py

KITTI Dataset

Download data and running

You can download the KITTI Dataset from http://www.cvlibs.net/datasets/kitti/raw_data.php

The directory structure is as follow

└─KITTI Dataset path
    ├─testing
    │  ├─calib
    │  ├─image_2
    │  └─velodyne
    └─training
        ├─calib
        ├─image_2
        ├─label_2
        └─velodyne

runnning

python kitti_vis_single.py --dataset <dataset_path> --file_index <file_index>

Performance

Waymo open Dataset

You can download the Waymo Open Dataset from https://waymo.com/open/data/perception/

Then you can visualize single tfrecord file as follow

python waymo_vis_sequence.py --filepath <tfrecode_file_path>

nuScense Dataset

You can download the nuScense Dataset from https://www.nuscenes.org/download

The directory structure is as follow

└─nuScense Dataset path
    ├─maps
    ├─samples
    ├─sweeps
    │  ├─LIDAR_TOP
    │  ├─...   
    ├─v1.0-trainval
    ...

runnning

python nuscense_vis_sequence.py --rootpath <nuScense_root_path>