1 of 1

Latent Representations for Particle Feature Exploration

Scientific Achievement

A new particle latent representation that combines geometric and attribute information to facilitate particle feature extraction and tracking.

Significance and Impact

The semi-automatic feature exploration and tracking are shown to be more effective than hand-craft descripts for features extraction

Technical Approach

Apply Geometric Convolution (GeoConv) to generate local latent vectors for particles where explicit connectivity is not available.
Perform hierarchical clustering and mean-shift tracking algorithm in latent space to achieve semi-automatic feature exploration and tracking

Feature exploration tool in the latent space

Extracted finger structures compared to topological analysis

Haoyu Li and Han-Wei Shen: Local Latent Representation based on Geometric Convolution for Particle Data Feature Exploration, IEEE Transactions on Visualization and Computer Graphics (2022)

Feature related particle data analysis plays an important role in many scientific applications such as fluid simulations, cosmology simulations and molecular dynamics.

Compared to conventional methods that use hand-crafted feature descriptors, some recent studies focus on transforming the data into a new latent space, where features are easier to be identified, compared and extracted.

However, it is challenging to transform particle data into latent representations, since the convolution neural networks used in prior studies require the data presented in regular grids.

We adopt Geometric Convolution, a neural network building block designed for 3D point clouds, to create latent representations for scientific particle data.

These latent representations capture both the particle positions and their physical attributes in the local neighborhood so that features can be extracted by clustering in the latent space and tracked by applying tracking algorithms such as mean-shift.

We validate the extracted features and tracking results from our approach using datasets from three applications and show that they are comparable to the methods that define hand-crafted features for each specific dataset.