Image-Dataset-Statistics

This repository provides an easily customizable script for the initial analysis of image datasets, specifically for use in machine learning.

Features

• Calculate and visualize brightness distributions across datasets
• Generate violin plots for dataset comparisons
• Perform statistical significance tests:
  • Mann-Whitney U test for comparing two datasets
  • Kruskal-Wallis test with post-hoc analysis for multiple datasets
• Support for both images and video files (common formats: jpg, png, mp4, avi)

Installation

git clone https://github.com/janikhohmann/Image-Dataset-Statistics.git
cd Image-Dataset-Statistics
pip install -r requirements.txt

Output

The tool generates:

• Violin plots showing brightness distributions
• Statistical test results including p-values
• CSV files with detailed statistics