Synthetic and scanned data of BridgeDiff

Description

To address the lack of large-scale point clouds dataset for training point cloud completion models in civil engineering, we publish this dataset. This dataset comprises the synthetic and case study datasets used in the paper ‘BridgeDiff: A two-stage conditional diffusion model for bridge point cloud completion’. This dataset comprises 1,000 synthetic single-span masonry arch bridge point cloud samples for point cloud completion. Each single-span sample includes 1 ground truth point cloud and 24 partial point clouds with varying perspectives and degrees of data loss, a total of 24,000 sample pairs for training. For ease of processing, each sample is stored as a binary (.pt) file. The two case studies are a canal bridge in Birmingham, UK, and the Welland Viaduct in Harringworth, UK. Both point clouds have undergone registration, denoising and downsampling, the full scan data will be made available once the paper has been accepted.

Institutions

• University of Birmingham

  England, Birmingham
• Durham University

  England, Durham

Categories

Licence