The input, synthetic training data and output of proposed point cloud denoising method

The data used in a Data-Centric Engineering paper.

Based on the model of Dr. Wei Lin, the data was generated and analyzed using the method proposed in the paper "Denoising image point clouds using segmentation and synthetic data for enhanced structural health analysis of tunnels."

The training dataset (2-1/2-2, 1-1~1-16) is organized in 5 columns: x, y, z, -, and flag. The predicted dataset (1-17~1-20) is organized in 7 columns: x, y, z, -, flag_syn, flag_pre, -.

In case the files are not correctly shown in folders after uploading in Zenodo: the data is organized in folds synthetic training data (2-1, 2-2, and 1-1~1-16, among which the bad dataset is deleted), log (checkpoint.pt, log_test.txt, and log_train.txt), output (1-17~1-20), file config.py for data generation, and files DPT12_0530_group1_densified_point_cloud - Cloud.subsampled.subsampled.subsampled.subsampled as the data processing result.

This study is funded by China Scholarship Council.