detailing readme.md

Pamphlett · Pamphlett · commit 1a0198c6be2b · 2023-01-18T20:22:05.000+08:00
diff --git a/README.md b/README.md
@@ -1,14 +1,21 @@
 # ***Segreagator***: Global Point Cloud Registration with Semantic and Geometric Cues
 
-We present Segregator, a global point cloud registration pipeline using both semantic and geometric information.
+Pengyu Yin, Shenghai Yuan, [Haozhi Cao](https://www.researchgate.net/profile/Haozhi-Cao), Xingyu Ji, Shuyang Zhang, and [Lihua Xie](https://dr.ntu.edu.sg/cris/rp/rp00784)
+[ICRA2023][preprint coming soon]
+
+We present Segregator, a global point cloud registration pipeline using both semantic and geometric information. Instead of focusing solely on point level features, we build degenerancy-robust correspondences between two LiDAR scans on a mixed-level (geometric features as well as semantic clusters). Additionally, G-TRIM based outlier pruning is also proposed to find out the inlier correspondence set more efficiently. Please refer to our paper for more details.
 
 ![](assets/segregator_intro.png)
 
+----
+
 ### Test Environment
 * Linux 18.04/20.04 LTS
 * ROS Melodic/Noetic
 
-### How to Build
+----
+
+### Installation
 Run the following lines for denpandencies:
 ```
 sudo apt install cmake libeigen3-dev libboost-all-dev
@@ -25,15 +32,34 @@ cd ~/catkin_ws
 catkin build segregator 
 ```
 
-### How to Run
-With the provided scans in the ```materials``` folder, run the following lines in the catkin workspace to reproduce the figure above:
+----
+
+### Test on different datasets
+* ##### Toy example on KITTI
+We include two distant scans from [KITTI](https://www.cvlibs.net/datasets/kitti/) dataset sequence 00 in the ```materials``` folder, run the following lines in the catkin workspace to reproduce the figure above:
 ```
 source devel/setup.bash
 roslaunch segregator run_segregator.launch
 ```
+* ##### On other/self-collected dataset
+Generally, apart from the pointcloud file itself, per-point semantic label is also needed to make Segregator work. We recommend using [SPVNAS](https://github.com/mit-han-lab/spvnas/blob/master/README.md#news) (the most accurate), [Rangenet](https://github.com/PRBonn/rangenet_lib) or [SalsaNext](https://github.com/TiagoCortinhal/SalsaNext) (far more computationally efficient, range image-based methods with a bit segmentation quality drop) to generate the labels.
+
+----
 
 ### Comparative Results
 TO BE ADDED
 
+----
+
+### Contact
+Please kindly reach out to me if you have any question. Discussions are also welcome:
+Pengyu Yin ([pengyu001@ntu.edu.sg]())
+
 ### Acknowledgements
-We would like to thank [Quatro](https://github.com/url-kaist/Quatro), [Teaser](https://github.com/MIT-SPARK/TEASER-plusplus) as well as [T-LOAM](https://github.com/zpw6106/tloam) for making their project public.
+This research is supported by the National Research Foundation, Singapore under its Medium Sized Center for Advanced Robotics Technology Innovation ([CARTIN](https://www.ntu.edu.sg/cartin)).
+
+Also, we would like to show our greatest thankfulness to the authors of the following repos for making their own works public:
+* [Teaser](https://github.com/MIT-SPARK/TEASER-plusplus) (baseline)
+* [Quatro](https://github.com/url-kaist/Quatro) (baseline)
+* [SPVNAS](https://github.com/mit-han-lab/spvnas) (semantic segmentation)
+* [T-LOAM](https://github.com/zpw6106/tloam) (clustering) 
