Quick Start

Put your raw data in right place

Put your raw point cloud data in ./data

File structure should be like this

.
└── 2-l
    ├── 2-l - Cloud-body.txt
    ├── 2-l - Cloud-foot1.txt
    ├── 2-l - Cloud-foot2.txt
    ├── 2-l - Cloud-foot3.txt
    ├── 2-l - Cloud-foot4.txt
    ├── 2-l - Cloud-head.txt
    ├── 2-l - Cloud-tail.txt
    ├── 2-l-1.JPG
    ├── 2-l-2.JPG
    ├── 2-l-3.JPG
    ├── 2-l-4.JPG
    ├── 2-l-5.JPG
    ├── 2-l-6.JPG
    └── 2-l.ply

2-l.ply is the raw point clouds

*.txt is the parts' segmentation for annotations

*.jpg is the different view of 3D object

Please make sure the file extension is .ply->raw point clouds, .txt->segmentated point cloud

You can download example from https://github-1253353217.cos.ap-beijing.myqcloud.com/data-2-l.zip

Preprocess the dataset

python main.py -s ./data -d ./res

The result will be stored in ./res

File structure of ./res

.
└── 2-l
    ├── clean_annotation.txt
    ├── clean_100k.ply
    ├── clean_10k.ply
    ├── clean_1k.ply
    ├── info.txt
    ├── mesh.ply
    ├── noiseless.ply
    ├── real_annotation.txt
    ├── real_100k.ply
    ├── real_10k.ply
    └── real_1k.ply

info.txt is basic info for this 3D shape

annotation.txt is info for annotations

clean_*k.ply is point cloud sampled from reconstructed mesh

real_*k.ply is point cloud sampled from noiseless point cloud

noiseless.ply is noiseless point cloud sampled from raw point cloud

You can download example from https://github-1253353217.cos.ap-beijing.myqcloud.com/res-2-l.zip

Example for clean_100k

Show annotation

python show_annotation.py -i ./res/2-l -t c # for clean_100k.ply
python show_annotation.py -i ./res/2-l -t r # for real_100k.ply

Example for annotation

Visual Checker (Recommended)

You can use visual_checker.py to check your point cloud and mesh.

python visual_checker.py -s res -d check.csv

程序使用说明:

    程序使用说明
    选中绘图窗口按快捷键：（关闭输入法）
    Q: 下一个
    A: 正常
    Z: 噪音过多 (手工难以剔除)
    X: 明显孔洞
    C: 车窗孔洞
    V: 明显缺失
    S: 切换分割视图 (有分割文件才可以)
    D: 分割标注异常 (如果仅是分割有孔洞，原模型没有孔洞，则记为分割标注异常，而非明显孔洞)
    W: 有离群点 (手工可以剔除)
    R: 类别不明
    E: 重建失真
    space: 切换背景 （绿色背景，更容易看出问题）
    .: 强制中断程序

Also, you can use GUI version for easy check

python visual_checker_GUI.py -s res -d check.csv