Neural Biodiversity (Sofia Crespo — Neural Zoo / Artificial Natural History)

Math essence

For each creature i: compute r(θ) = r_i·(1 + Σ_k a_{i,k}·sin(k·θ + φ_{i,k})); mask m_i = clip(1 - (r(x,y) - r(θ))/δ_i, 0, 1); texture t_i = sin(k_x·x + k_y·y + ψ_i); colour c_i = palette_i · (0.55 + 0.55·t_i); composite canvas ← canvas·(1-α·m_i) + c_i·α·m_i.

Parameters

Reference

First 60 lines of reference.py. Full source: library_external/ml_artists/CrespoNeuralBiodiversity/reference.py

"""CrespoNeuralBiodiversity — Sofia Crespo "Neural Biodiversity" kernel.

Sofia Crespo's *Neural Zoo* (2018-), *Artificial Natural History*
(2019-), *Artificial Remnants* (2020-) and *Beneath the Neural Waves*
(2021) use 3D GANs, CNN-based texture synthesis and 3D style transfer
to fabricate a "natural history book that never was": chimerical
creatures whose anatomies recombine fragments of real species.  Her
neural networks learn *texture*, not shape — the silhouette of each
creature emerges from overlapping fields of organic-looking surface
patterns.

We replicate the kernel in pure numpy:
  1. Place N "creature sites" at random positions on the canvas.
  2. Each creature is a layered field of sinusoidal blobs:
     concentric rings + radial lobes + a slowly-varying texture
     derived from layered sine functions of (r, θ).
  3. Render each creature as a soft coloured disk with a sinusoidal
     outline perturbation — the outline is r(θ) = r0 · (1 +
     ε · Σ_k sin(k·θ + φ_k)).
  4. Compose all creatures into a single biodiversity field with
     semi-transparent overlap so layered forms suggest anatomy.

References:
- https://www.thisiscolossal.com/2020/09/sofia-crespo-ai-natural-history/
- https://sofiacrespo.com/neural-zoo/
- https://www.nvidia.com/en-us/research/ai-art-gallery/artists/sofia-crespo/
- https://www.ted.com/talks/sofia_crespo_ai_generated_creatures_that_stretch_the_boundaries_of_imagination
"""
from __future__ import annotations

from typing import ClassVar

import numpy as np


class CrespoNeuralBiodiversity:
    """A field of layered sinusoidal creatures (Neural Biodiversity).

    Parameters
    ----------
    dim : int
        Square canvas side length (px).
    n_creatures : int
        Number N of creature sites.
    base_radius : float
        Average creature radius (px).
    radius_jitter : float
        Per-creature multiplicative radius jitter (0..1).
    outline_amp : float
        Sinusoidal outline perturbation amplitude (fraction of r0).
    outline_freq : int
        Number of lobes in the outline sinusoid (typical 3-8).
    texture_freq : float
        Spatial frequency of internal texture pattern.
    palette : str
        ``"crespo_coral"``, ``"crespo_insect"``, ``"crespo_anemone"``
        or ``"random"``.
    """

    family: ClassVar[str] = "ml_artists"