This page organizes the datasets, 3D models, and reproducible code repositories that support my dissertation research on carnivore bone surface modifications (BSMs). Repositories are grouped by workflow stage: (1) specimen metadata, (2) 3D scans and processing protocols, and (3) analysis code for landmarking, variable extraction, and statistical modeling.
Specimen resources are organized around two complementary datasets: curated museum collections used to build comparative reference models, and experimental assemblages used to test equifinality and validate measurement and classification workflows under controlled conditions.
Comparative reference specimens scanned and analyzed during NMNH collections work.
Experimental datasets and teaching/research assemblages used for method development and robustness testing. Collections are located at the University of Wisconsin- Madison (Bunn), Southern Connecticut State University (Selvaggio), and University of Arkansas (Terhune).
Typical tools: structured metadata tables (CSV), Git/GitHub, and consistent folder conventions for reproducibility.
Scan repositories include processed surface meshes used for analysis, along with documentation for acquisition, post-processing, and quality control. Protocols emphasize consistent preprocessing so that curvature-based and landmark-based measures are comparable across specimens and instruments.
Cleaned and standardized 3D surface meshes (e.g., PLY/OBJ) prepared for morphometrics and surface topography. Where applicable, repositories include decimated or permission-approved derivatives rather than raw archival scans.
Step-by-step documentation for scanning, mesh cleaning, remeshing/decimation, and coordinate conventions. Includes QA/QC checks (normals, face counts, topology flags) and batch-processing guidance for consistent outputs.
Typical tools: David SLS-3, Artec Spider II, Transcan C, Geomagic Design X, MeshLab/Blender, and standardized mesh QA checks.
Automated landmarking utilities designed for high-throughput 3D mesh workflows. The goal is to reduce manual bottlenecks while keeping placements reproducible, inspectable, and compatible with downstream R morphometrics.
Scripts/macros to place fixed landmarks consistently across many specimens in 3D Slicer, export standardized landmark files, and log run metadata. Supports batch processing and minimizes manual steps while preserving anatomical landmark logic.
Quick visual checks to confirm landmark order, sidedness, and surface projection. Includes examples for overlaying landmarks on meshes, generating QA snapshots, and flagging specimens that need manual review.
Typical tools: 3D Slicer (Python), VTK-based workflows, mesh IO (PLY/OBJ), Git/GitHub, and lightweight QA images/logs.
Variable-extraction pipelines that complement landmark-based morphometrics. These workflows produce modeling-ready tables and include preprocessing checks so that derived variables can be traced back to input meshes and parameters.
Scripts for generating additional shape descriptors beyond fixed landmarks, including outline-based approaches such as Elliptic Fourier Analysis. Useful for exploratory comparisons, dimensionality reduction, and feature integration alongside other variables.
Batch workflows for surface-derived metrics (e.g., curvature-based measures) and associated QA/QC steps. Includes fallbacks for cleaning and decimation when meshes exceed face-count thresholds or contain minor topology issues.
Typical tools: geomorph, Morpho, tidyverse, batch file IO, and reproducible export tables (CSV) for modeling.
Statistical and machine-learning workflows used to evaluate group structure, quantify separation among classes, and predict mark type and/or taxon actor from combined morphometric and surface-derived variables.
Reproducible scripts for ordination, group separation, and model interpretation. Includes publication-quality plots (e.g., PCA/CVA scatterplots, deformation/TPS visualizations where applicable) and embedded notes describing what each output indicates.
Predictive modeling pipelines built around transparent evaluation: train/test splits, cross-validation, feature selection, and error interpretation (confusion matrices, class imbalance checks, sensitivity analyses). Designed to support robust inference before applying models to fossil datasets.
Typical tools: geomorph, tidyverse, ggplot2, caret/tidymodels (or equivalent), and consistent session/documentation practices.