Walk-through-Raster (Nake)

Math essence

Given a chain position t ∈ [0,1] we set the base distribution p(sign | t) with p_vert ∝ (1−|2t−1|) and p_horiz ∝ (0.4+0.6·(1−|2t−1|)), then bias toward the previous sign (vertical sticks to vertical) before sampling. Each chosen sign is rasterized into its cell at the (ci, ri) index produced by the chosen mapping mode.

Parameters

Reference

First 60 lines of reference.py. Full source: library_external/generative_pioneers/nake_walk_sequence/reference.py

from __future__ import annotations

from typing import ClassVar

import imageio.v2 as imageio
import numpy as np


class NakeWalkSequence:
    family: ClassVar[str] = "generative_pioneers"
    sketch: ClassVar[str] = "nake_walk_sequence"

    def __init__(
        self,
        dim: int = 512,
        cols: int = 28,
        rows: int = 36,
        transition_seed: int = 11,
        p_vertical_peak: float = 0.55,
        p_horizontal_peak: float = 0.18,
        fade_alpha: float = 0.30,
        bg_color: tuple[int, int, int] = (245, 240, 230),
        ink_color: tuple[int, int, int] = (15, 15, 25),
        accent_color: tuple[int, int, int] = (200, 60, 50),
        mapping_mode: str = "row",
        seed: int | None = None,
    ) -> None:
        self.dim = dim
        self.cols = cols
        self.rows = rows
        self.transition_seed = transition_seed
        self.p_vertical_peak = p_vertical_peak
        self.p_horizontal_peak = p_horizontal_peak
        self.fade_alpha = fade_alpha
        self.bg_color = np.array(bg_color, dtype=np.uint8)
        self.ink_color = np.array(ink_color, dtype=np.uint8)
        self.accent_color = np.array(accent_color, dtype=np.uint8)
        self.mapping_mode = mapping_mode
        self._seed = seed
        self.rng = np.random.default_rng(seed)
        self.canvas = np.full((dim, dim, 3), self.bg_color, dtype=np.uint8)
        self.time = 0
        self._dirty = True

    def reset(self, seed: int | None = None) -> None:
        if seed is not None:
            self._seed = seed
        self.rng = np.random.default_rng(self._seed)
        self.canvas = np.full((self.dim, self.dim, 3), self.bg_color, dtype=np.uint8)
        self.time = 0
        self._dirty = True

    def _transition_probs(self, position_t: float) -> dict:
        center = 1.0 - abs(2.0 * position_t - 1.0)
        p_vert = self.p_vertical_peak * center
        p_horiz = self.p_horizontal_peak * (0.4 + 0.6 * center)
        p_dot = 0.04
        p_empty = max(0.05, 1.0 - p_vert - p_horiz - p_dot)
        total = p_empty + p_vert + p_horiz + p_dot
        return {