LiDAR in Soccer Processing Team

This reppository is for processig point clouds data of LiDAR in soccer.

Data structure

Please edit settings.py file and change DATA_PATH and BASE_PATH to yours.
Data structure under DATA_PATH is as follows.

DATA_PATH
├─LIVOX_Hallway
|  └─jogging_fast_4th.lvx
├─LIVOX_Hallway_csv
|  └─jogging_fast_4th.csv
├─LIVOX_Hallway_pcds
│  └─jogging_fast_4th
│      └─res10ms_start13.5s
├─New_data
│  └─New_data
│      └─pcds
└─recorded data
    └─recorded data

We first focus on jogging_fast_4th.lvx data.

Get frames from csv data

At first, get jogging_fast_4th.csv file from livox viewer(use Tools > File Convertor > lvx to csv).

input: csv file
output: sequence of pcd files
You can change time resolution per frame and timing of frame0 by using res and start arguement.
Use csv2pcd.ipynb or run get_frames/csv2pcd.py

Clustrization

input: raw pcd files
output: clusterized labels
By using check_clusterization.ipynb, you can clusterize point clouds and visualize the clusters.

At the first cell, you can process sequential pcd files, and second, single pcd file. At the first cell, you firstly select the viewpoint and close the window. After that, you can see the clusters from the selected viewpoint.
You can process single pcd files also by the clusterization/clusterize.py, but not sequential data.

In both cells, you can change data_path. And, at the clusterization.clusterize function, you can define more arguments than written in .ipynb file. So, please take a look at clusterization/clusterize.py.

Human detection

Compute CoM(center of mass)

Evaluation