Update README.md

README.md

@@ -44,7 +44,7 @@ For the two dataloaders of datasets, we provide dataloader scripts in the `datas
 ## Custom Data Processing
 For self-processing data, please check the scripts in 'data_preprocess' folder for each individual training data processing:
 If you own dataset is ordered sequence point cloud frames, just reuse the same KITTI processing script to process the sequentail point cloud frames, So the source and target scans are using $i$ th and $i+1$ th frame respectively.
-- `process_KITTI_Feature.py`: For processing sequential dataset
+- `process_KITTI_Feature.py`: For processing sequential dataset.<br> 
 For processing KITTI dataset. Otherwise, if your point cloud frames are unordered, please refer to the 3D Match script to process, yet you have to establish the correspondence between source and target point scans, with a minimum 30% point overlapping between source and target scans, otherwise, we refer you to use public library like Open3D, PCL (Point Cloud Library), scikit-learn, through KDTree or Octree to create source and target frame correspondence with engouh point overlappings. Original 3DMatch already processed it for use. For further scan pair match, you can refer to the PointDSC repository to process the feature descriptors, [FPFH](https://github.com/XuyangBai/PointDSC/blob/master/misc/cal_fpfh.py), [FCGF](https://github.com/XuyangBai/PointDSC/blob/master/misc/cal_fcgf.py), as most of our data preprocessing codes are adapted based on their codes. 
 - `process_3DMatch_Feature.py`: For processing paired scan dataset with enough overlapping > 30%.