Visual Acoustics Open Dataset: Cross-modal correspondence metrics with human perceptual validation

This repository contains a complete computational pipeline for the analysis of iris images, developed as part of the Argira Station research project (Visual Acoustics). The pipeline integrates: - **Image preprocessing** (CLAHE enhancement, Gaussian blur)- **Automatic pupil/iris detection** using Hough circle transforms and adaptive refinement- **Radial and angular profile extraction** to capture intensity variations- **Angular FFT analysis** to identify dominant periodicities (crypts, furrows, pigment patterns)- **Golden ratio (φ) comparison** of the iris/pupil radius ratio- **Fractal dimension estimation** via box‑counting on edge‑detected structures- **Sonification** of the angular profile using additive synthesis (φ‑scaled frequencies)- **Validation** with synthetic iris images containing ground‑truth parameters The work extends the Visual Acoustics open research platform ([DOI: 10.5281/zenodo.18734732](https://doi.org/10.5281/zenodo.18734732)) with applications in ophthalmology, biometrics, and generative art. --- ## 📁 Files included | File | Description ||------|-------------|| `iris_analysis.py` | Main pipeline for real iris images: detection, profile extraction, FFT, fractal dimension, sonification, and figure generation. || `iris_synthetic_v2.py` | Synthetic iris generator with controlled ground‑truth parameters (pupil/iris radii, angular peaks). Used for validation and robustness testing. || `iris_analysis_real.png` | Summary figure of the analysis on a sample real iris. || `iris_synthetic.png` | Example of a synthetic iris image generated by `iris_synthetic_v2.py`. || `iris_validation.png` | Validation figure comparing detection vs ground‑truth under different noise levels. || `iris_sound.wav` | Audio sonification of the analysed iris (generated by the pipeline). || `README.md` | This file. | Usage 1. Analysing a real iris image Step 1: Prepare your image Place a frontal iris image (e.g., my_iris.jpg) in the working directory. Good quality images with visible iris texture work best. Step 2: Edit the script Open iris_analysis.py and modify the IMG_PATH variable (line ~30) to point to your image: python IMG_PATH = '/path/to/your/iris.jpg' ## 🚀 Usage ### 1. Analysing a real iris image #### Step 1: Prepare your image Place a frontal iris image (e.g., `my_iris.jpg`) in the working directory. Good quality images with visible iris texture work best. #### Step 2: Edit the script Open `iris_analysis.py` and modify the `IMG_PATH` variable (line ~30) to point to your image: ```python IMG_PATH = '/path/to/your/iris.jpg' ``` Optionally adjust parameters: · max_dim (line 34): resize image for faster processing (default 900px) · duration (line 420): length of sonification in seconds (default 8.0) · n_angles (line 146): angular resolution (default 360) Step 3: Run the analysis ```bash python iris_analysis.py ``` Step 4: Outputs The script generates: · Terminal output: detection results, metrics, fractal dimensions, FFT peaks · iris_analysis.png: summary figure with 7 panels (detection, edges, radial/angular profiles, FFT, sonification waveform, polar map) · iris_sound.wav: 8‑second audio sonification --- 2. Generating and validating with synthetic irises Run the synthetic pipeline ```bash python iris_synthetic_v2.py ``` This script: · Generates a synthetic iris with known parameters: · Pupil/iris radii (φ‑based ratio) · Embedded angular frequencies (e.g., peaks at k = 5, 8, 13) · Adjustable noise levels (σ = 2, 8, 20, 40) · Runs the detection algorithm on each noisy version · Compares detected vs ground‑truth values · Produces iris_validation.png (8‑panel figure showing detection under noise) Customising the synthetic iris Inside iris_synthetic_v2.py, you can modify: · size (default 600): image dimensions · fib_peaks (default [5,8,13]): angular frequencies to embed · noise_std (default 8.0): baseline noise level · noise_levels (line 280): list of σ for robustness tests --- 3. Understanding the sonification The sonification maps angular FFT peaks to audible frequencies using: ``` freq_hz = 220 * φ^(k/6) ``` where k is the angular frequency (cycles per revolution) and φ ≈ 1.618. This creates a harmonic series where each partial is scaled by the golden ratio. The amplitude of each partial is modulated by the angular profile intensity, and a slow envelope fades in/out to avoid clicks. 4. Using the interactive web interface Open index.html in any modern browser to explore the sonification process in real time. The interface allows you to: · Upload images and see their visual metrics extracted instantly. · Listen to the generated sonification. · Rate the result (1–5 stars) to contribute anonymous data to the public dataset. · View the collected dataset and download it as JSON. No image is ever uploaded to any server — all processing happens locally in your browser. --- 📊 Expected outputs (iris analysis) (I will provide the image once the user has given written consent.) For a real iris (example from iris_analysis_real.png) Metric Value Interpretation Pupil radius 255 px – Iris radius 908 px – Iris/pupil ratio 0.281 Compared to φ⁻¹ ≈ 0.618 (deviation 54%) Pigment coverage 48.1% Sectorial amber/brown heterochromia Fractal dimension (edges) 1.2705 Moderate complexity, organised tissue Fractal dimension (pigment) 1.7790 Close to DLA dimension (~1.8), suggesting fractal aggregation Angular FFT peaks k = 5, 8, 13, … Fibonacci‑related periodicities For synthetic validation The pipeline should recover embedded frequencies with high accuracy at low noise (σ ≤ 8), and maintain acceptable performance up to σ = 20. --- 📝 Citation If you use this work, please cite: Ranero García, J. (2026). Visual Acoustics Open Dataset: Cross-modal correspondence metrics with human perceptual validation (Version v4). Zenodo. https://doi.org/10.5281/zenodo.18763683 For the foundational work on phyllotaxis: Ranero García, J. (2026). Computational Verification of Vogel's Model in Phyllotaxis Patterns. Zenodo. DOI: 10.5281/zenodo.18714818 --- 🔗 Related projects · K‑nacci Solar Layouts v7 – Optimisation of solar panel arrangements using phyllotaxis and discrepancy theory. · Visual Acoustics (previous version) – Earlier release of the platform. --- 📜 License This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0). You are free to share and adapt the material for any purpose, even commercially, as long as you give appropriate credit. -- 📬 Contact Jose Ranero García – Independent Researcher 🌌 Argira Station 📡 YouTube Channel ☄️🔭📡 Last updated: February 2026