FractalPerception.com

Exploring Aesthetic Perception in High-Dimensional Parametric Fractal Systems

1. The Project

FractalPerception is an independent research program aimed at exploring aesthetic perception within visual environments generated from parametric fractal systems.

It establishes a structured correspondence between parametric variation in a mathematical space and visual perception, based on a corpus of high-complexity, non-natural images.

Unlike conventional image datasets composed of static pixels, this corpus provides access to the underlying generative structures (parameters, layers, transformations), while accounting for the fact that the final rendering depends on a partially non-observable process.

The images produced do not belong to a figurative representational regime and do not refer to identifiable objects. They constitute a non-referential visual environment, allowing perception to be studied independently of recognition mechanisms and cultural biases.

This corpus is not designed as a dataset generated on demand, but as the result of an ongoing artistic and research practice, which defines its coherence and value.

2. Mathematical Structure of the Images

All visual productions are generated exclusively from parametric fractal systems. Each image is described by:

• a fractal formula family
• a complete set of numerical variables
• spatial transformations (zoom, position, rotation)
• a color palette or associated color system
• multi-layer compositional parameters

This parametric description enables analysis beyond pixel space and provides direct access to the relationship between generative variation and visual structure.

However, this formalization does not guarantee strict reproducibility of the rendering. The generation pipeline introduces a non-observable component related to rendering engines and partially opaque internal processes.

This dissociation between parametric description and actual rendering provides a framework to study the gap between generative definition and perceived visual output.

Each image is also associated with a semantic annotation derived from the author’s interpretation, introducing an additional dimension to analyze relationships between structure, aesthetic selection, and perceived meaning.

3. The Corpus

The corpus is continuously enriched through Philippe Chevalier’s ongoing artistic research.

It currently includes:

• 116 unique functions
• up to 231 occurrences for a single function
• 1,020 parameterized layers, with an average of 21 parameters per layer
• 21,569 parameters at the layer level (excluding gradient and opacity parameters)
• 376 fractal compositions, including 135 considered finalized, with an average of 91 parameters per image
• 34,369 parameters at the composition level (excluding gradient and opacity parameters)
• 1,396 images with full mathematical descriptions
• a very high-resolution bitmap rendering for each image

This corpus constitutes a structured space for exploring non-natural visual distributions characterized by organized and controlled complexity.

4. Perspective

In the short term, the corpus can be used as an instrument to explore:

• the relationship between generative parameters and visual rendering
• perceptual invariants in complex structures
• model behavior on non-natural distributions

For both research teams and organizations developing generative systems, this corpus provides a structured and actionable anchor point, where most available data remains unstructured and non-parametric.

In the mid term, the project may integrate a process capture system, recording exploration trajectories, successive adjustments, and selection mechanisms. This would enable the construction of a dynamic corpus integrating temporal and decision-making dimensions, opening the study of exploration processes and the formation of aesthetic judgment.

In the longer term, extending the system to a broader range of observers could enrich the data and support the development of a large-scale corpus of aesthetic perception.

5. Collaboration

Philippe Chevalier develops an independent research practice at the intersection of parametric generative systems, aesthetic perception, and the production of high-complexity non-natural images.

The FractalPerception corpus is not designed as a dataset produced on demand, but as the outcome of an ongoing artistic and research process. Its autonomy, internal coherence, and partial non-reproducibility are precisely what give it experimental value.

Within this framework, FractalPerception is used as a working instrument to explore questions related to:

perceptual alignment and aesthetic coherence in complex visual structures
model robustness and behavior on non-natural distributions
the gap between parametric generative description and perceived visual output

Philippe Chevalier collaborates with research and R&D teams as a consultant, providing access to this evolving corpus and contributing to its interpretation in specific experimental contexts.

He is the author and architect of the corpus — its design, parametric structure, and aesthetic selection process cannot be dissociated from the practice that produces it.

Collaborations may take the form of joint explorations, targeted analyses, or integration of the corpus into experimental protocols.

6. Artistic Research

FractalPerception is part of Philippe Chevalier’s broader artistic research on parametric fractal systems.

The complete body of finalized works and the overall development of this research can be viewed at: www.philoxerax.com