Replication Data for: Making clustering methods workable for shapes using the ordinary Procrustes sum of squares

Description

This dataset contains the R programs and preprocessed text datasets used in the experiments for the paper "Making clustering methods workable for shapes using the ordinary Procrustes sum of squares" (Pattern Recognition, 2026, DOI: 10.1016/j.patcog.2025.111878). It includes implementations of various shape clustering methods, including K-means, fuzzy c-means, spectral clustering, mean shift, and convex clustering. Bundled preprocessed text datasets include Skewed line drawings (SLD), NFL Big Data Bowl 2020 subsets, and MLB Scherzer's pitch 2021 subsets.

Steps to reproduce

1. Install the required R packages: `ggplot2`, `ggrepel`, `tidyr`, `gridExtra`, `RColorBrewer`, `plotly`, and `shapes`. 2. Set the working directory in R to the root of this dataset. 3. To run a quick sanity check, execute: source("shapeclusts.R") shapeclust.kmeans("drawings", 3, ndim = 2, ntrial = 5, dist = "oss") 4. Check the generated output files (e.g., joule.txt, ev.txt) in the working directory. 5. Refer to the "Used commands in the experiments" section in the README.md file to reproduce the full experimental results.

Institutions

• Hiroshima City University

  Hiroshima, Hiroshima

Categories

Funders

• Japan Society for the Promotion of Science

  Ministry of Education, Culture, Sports, Science and Technology

  Tokyo

  Grant ID: 22K12171

Related Links

Licence