Skip to content

This is a repository mainly about IEEE data fusion contest 2019 track 4 — Cloud points classification

Notifications You must be signed in to change notification settings

ShoupingShan/dfc2019_track4

Repository files navigation

IGRSS Data Fusion Contest 2019 point cloud segmentation track rank 3 method

Data description

Track 4 is the 3D Point Cloud Classification track. The goal is to classify (semantically segment) point clouds on a per point basis. The classes are:

Class Index Class Description
2 Ground
5 High Vegetation
6 Building
9 Water
17 Bridge Deck

Baseline

For the baseline algorithm, a PointNet++ model was updated with modifications to support splitting/recombining large scenes.

PointSIFT is a semantic segmentation framework for 3D point clouds. It is based on a simple module which extract featrues from neighbor points in eight directions.

Result

Matrix

Confusion matrix with overall accuracy: 98.33%
T P(2) P(5) P(6) P(9) P(17) 
---------------------------------------------------
2 5103290| 65| 20517| 4408| 3066|
5 1810| 1343702| 28977| 2| 4|
6 47554| 16342| 1322237| 0| 1036|
9 3379| 0| 0| 191191| 20|
17 5130| 0| 692| 0| 83878|

mIoU: 0.944256680661

IoU:
Class 2 ( Ground ): 0.9833
Class 5 ( High Vegetation ): 0.9655
Class 6 ( Building ): 0.9184
Class 9 ( Water ): 0.9605
Class 17 ( Elevated Road ): 0.8936

rank

Original repository

@pubgeo

Usage

Data Augumentation

$ cd utils
$ please refer to utils/README.md

Date Preparation

$ cd dfc
$ python create_train_dataset.py --help

Train & Eval

$ chmod 777 run_5_fold.sh
$ ./run_5_fold.sh

Visualization

We have provided a handy point cloud visualization tool under utils. Run sh compile_render_balls_so.sh to compile it and then you can try the demo with python show3d_balls.py.

Co-Author

@ShoupingShan

@yinianqingzhi

Contact us

shp395210@outlook.com

About

This is a repository mainly about IEEE data fusion contest 2019 track 4 — Cloud points classification

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published