K-Means Analysis Minimal

A minimal project to run k-means split analysis with metadata.

Files

• explore_splits.py - Main analysis script
• splits/ - Split data files
  • k_means_split.json - K-means clustering splits
  • intersect_attributes.json - Available attributes list
• data/ - Required data files
  • all_filepath.txt - Image file paths
  • labels_images.txt - Image labels
  • mcrae-norms-grouped-with-concepts.json - Feature to concepts mapping
  • things/ - Concept metadata
    • concepts-and-categories.json - Concept definitions and categories
    • Categories_final_20200131.tsv - Supercategory mappings

Installation

pip install -r requirements.txt

Usage

python explore_splits.py

This will analyze the "has_wheels" attribute in the k-means splits and show:

• Split statistics (train/test distribution)
• Positive/negative sample counts
• Sample examples with metadata
• Supercategory distribution
• Summary statistics

Dependencies

• numpy
• pandas
• Standard Python libraries (json, csv, pathlib, collections, typing)