complexity-in-complexity

Code for "Complexity in Complexity: Understanding Visual Complexity Through Structure, Color, and Surprise". CogSci, 2025 & ICLR Re-Align Workshop, 2025.

https://github.com/complexity-project/complexity-in-complexity

Science Score: 54.0%

This score indicates how likely this project is to be science-related based on various indicators:

✓
CITATION.cff file
Found CITATION.cff file
✓
codemeta.json file
Found codemeta.json file
✓
.zenodo.json file
Found .zenodo.json file
○
DOI references
✓
Academic publication links
Links to: arxiv.org
○
Academic email domains
○
Institutional organization owner
○
JOSS paper metadata
○
Scientific vocabulary similarity
Low similarity (13.3%) to scientific vocabulary

Keywords

cognitive-science computer-vision image-processing python visual-complexity

Last synced: 6 months ago · JSON representation ·

Repository

Code for "Complexity in Complexity: Understanding Visual Complexity Through Structure, Color, and Surprise". CogSci, 2025 & ICLR Re-Align Workshop, 2025.

Basic Info

Host: GitHub
Owner: Complexity-Project
Language: Jupyter Notebook
Default Branch: main
Homepage: https://arxiv.org/abs/2501.15890
Size: 31.9 MB

Statistics

Stars: 0
Watchers: 0
Forks: 0
Open Issues: 0
Releases: 0

Topics

cognitive-science computer-vision image-processing python visual-complexity

Created about 1 year ago · Last pushed 9 months ago

Metadata Files

Readme Citation

Complexity in Complexity: Understanding Visual Complexity Through Structure, Color, and Surprise

This repository contains code, data, and scripts for the paper:

Complexity in Complexity: Understanding Visual Complexity Through Structure, Color, and Surprise
Karahan Sarıtaş, Peter Dayan, Tingke Shen, Surabhi S. Nath
arXiv preprint arXiv:2501.15890, 2025

Overview

We present novel interpretable features for visual complexity, combining:

Multi-Scale Sobel Gradient (MSG) to capture the patch-level symmetry at multiple scales,
Multi-Scale Unique Color (MUC) to quantify the colorfulness at multiple scales,
Surprise Scores derived from Large Language Models, indicating unusual/surprising objects or contexts.

Using these features alongside existing segmentation/object-based features, we demonstrate improved performance in predicting human-rated visual complexity across multiple datasets.

Quick Start

So, do you want to reproduce our results? Here's how you can do it in a few simple steps: 1. Clone and Install

bash git clone https://github.com/Complexity-Project/Complexity-in-Complexity.git cd Complexity-in-Complexity pip install -r requirements.txt

Run Experiments

Navigate to the linear/analysis.ipynb notebook and execute the first few cells to reproduce the results from the paper, and that's it! This will display the correlations between the regressors and human complexity ratings across all datasets.

Download & Unzip SVG Dataset

We further introduce a new dataset called Surprising Visual Genome (SVG) with surprising images from well-studied Visual Genome dataset along with human complexity ratings, to highlight the role of surprise in complexity judgments. We make this dataset publicly available for further research. Don't forget to cite our paper if you use this dataset in your research.

bash # Make sure you're inside the Complexity-in-Complexity/ directory # Then unzip the dataset into the "SVG" folder: unzip SVG_dataset.zip -d SVG

🤗 Alternatively, you can use our dataset directly from HuggingFace: https://huggingface.co/datasets/Mortdecai/SVG

Owner

Name: Complexity Project
Login: Complexity-Project
Kind: organization
Location: Germany

Repositories: 1
Profile: https://github.com/Complexity-Project

Citation (CITATION.cff)

cff-version: 1.2.0
message: "If you use our results/dataset/features/evaluation scripts, please cite the paper as below."
authors:
  - family-names: "Sarıtaş"
    given-names: "Karahan"
  - family-names: "Dayan"
    given-names: "Peter"
  - family-names: "Shen"
    given-names: "Tingke"
  - family-names: "Nath"
    given-names: "Surabhi S"
title: "Complexity in Complexity: Understanding Visual Complexity Through Structure, Color, and Surprise"
date-released: 2025-05-07
url: "https://arxiv.org/abs/2501.15890"
preferred-citation:
  type: article
  authors:
    - family-names: "Sarıtaş"
      given-names: "Karahan"
    - family-names: "Dayan"
      given-names: "Peter"
    - family-names: "Shen"
      given-names: "Tingke"
    - family-names: "Nath"
      given-names: "Surabhi S"
  year: 2025
  title: "Complexity in Complexity: Understanding Visual Complexity Through Structure, Color, and Surprise"
  url: "https://arxiv.org/abs/2501.15890"
  archivePrefix: "arXiv"
  eprint: "2501.15890"

ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Open Source Science