Diverse Topologies for Evaluation of Geometric Similarity Metrics: 7 Datasets of 3D Shapes

A collection of 7 datasets containing 3D shapes with varying topological complexity, created by Nivesh Dommaraju at the Technical University of Munich. The datasets include surface mesh and point cloud files and were used to compare metrics of geometric dissimilarity in a related journal article. Two of the datasets contain topologically complex shapes resembling designs from topology optimization.

Use Cases

• Evaluating geometric dissimilarity metrics based on 3D shape topology.
• Benchmarking clustering algorithms for unsupervised design exploration using labeled shape classes.
• Training or testing feature extraction methods like Pointcloud Autoencoders on synthetic, topologically complex designs.
• Comparing derived geometric feature metrics against common 3D object classification metrics.

Strengths

• Contains 7 distinct datasets, providing variety in topological complexity.
• Includes two datasets with shapes specifically resembling outputs from topology optimization, a key engineering method.
• Data is provided in both surface mesh (.stl) and point cloud (.ply) formats, offering flexibility for different processing pipelines.
• Some datasets are labeled with classes, enabling supervised or semi-supervised evaluation tasks.

Limitations

• The total number of designs across all datasets is not specified, limiting scale assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Technical University of Munich

Collection Method

Created for a related journal article on evaluating geometric similarity metrics.