#!/usr/bin/env python
-import math
+# Any copyright is dedicated to the Public Domain.
+# https://creativecommons.org/publicdomain/zero/1.0/
+
+# Written by Francois Fleuret <francois@fleuret.org>
+
+import math, sys, tqdm
import torch, torchvision
from torch import nn
from torch.nn import functional as F
-import cairo
-
-
-class Box:
- def __init__(self, x, y, w, h, r, g, b):
- self.x = x
- self.y = y
- self.w = w
- self.h = h
- self.r = r
- self.g = g
- self.b = b
-
- def collision(self, scene):
- for c in scene:
- if (
- self is not c
- and max(self.x, c.x) <= min(self.x + self.w, c.x + c.w)
- and max(self.y, c.y) <= min(self.y + self.h, c.y + c.h)
- ):
- return True
- return False
-
-
-def scene2tensor(xh, yh, scene, size=64):
- width, height = size, size
- pixel_map = torch.ByteTensor(width, height, 4).fill_(255)
- data = pixel_map.numpy()
- surface = cairo.ImageSurface.create_for_data(
- data, cairo.FORMAT_ARGB32, width, height
- )
- ctx = cairo.Context(surface)
- ctx.set_fill_rule(cairo.FILL_RULE_EVEN_ODD)
-
- for b in scene:
- ctx.move_to(b.x * size, b.y * size)
- ctx.rel_line_to(b.w * size, 0)
- ctx.rel_line_to(0, b.h * size)
- ctx.rel_line_to(-b.w * size, 0)
- ctx.close_path()
- ctx.set_source_rgba(b.r, b.g, b.b, 1.0)
- ctx.fill()
-
- hs = size * 0.1
- ctx.set_source_rgba(0.0, 0.0, 0.0, 1.0)
- ctx.move_to(xh * size - hs / 2, yh * size - hs / 2)
- ctx.rel_line_to(hs, 0)
- ctx.rel_line_to(0, hs)
- ctx.rel_line_to(-hs, 0)
- ctx.close_path()
- ctx.fill()
-
- return pixel_map[None, :, :, :3].flip(-1).permute(0, 3, 1, 2).float() / 255
-
-
-def random_scene():
- scene = []
- colors = [
- (1.00, 0.00, 0.00),
- (0.00, 1.00, 0.00),
- (0.00, 0.00, 1.00),
- (1.00, 1.00, 0.00),
- (0.75, 0.75, 0.75),
- ]
+######################################################################
- for k in range(10):
- wh = torch.rand(2) * 0.2 + 0.2
- xy = torch.rand(2) * (1 - wh)
- c = colors[torch.randint(len(colors), (1,))]
- b = Box(
- xy[0].item(), xy[1].item(), wh[0].item(), wh[1].item(), c[0], c[1], c[2]
- )
- if not b.collision(scene):
- scene.append(b)
-
- return scene
-
-
-def sequence(nb_steps=10, all_frames=False):
- delta = 0.1
- effects = [
- (False, 0, 0),
- (False, delta, 0),
- (False, 0, delta),
- (False, -delta, 0),
- (False, 0, -delta),
- (True, delta, 0),
- (True, 0, delta),
- (True, -delta, 0),
- (True, 0, -delta),
+
+colors = torch.tensor(
+ [
+ [255, 255, 255],
+ [255, 0, 0],
+ [0, 128, 0],
+ [0, 0, 255],
+ [255, 200, 0],
+ [192, 192, 192],
]
+)
+
+token_background = 0
+first_fish_token = 1
+nb_fish_tokens = len(colors) - 1
+token_forward = first_fish_token + nb_fish_tokens
+token_backward = token_forward + 1
+
+token2char = "_" + "".join([str(n) for n in range(len(colors) - 1)]) + "><"
+
+
+def generate(
+ nb,
+ height,
+ width,
+ max_nb_obj=2,
+ nb_iterations=2,
+):
+ pairs = []
+
+ for n in tqdm.tqdm(range(nb), dynamic_ncols=True, desc="world generation"):
+ f_start = torch.zeros(height, width, dtype=torch.int64)
+ f_end = torch.zeros(height, width, dtype=torch.int64)
+ n = torch.arange(f_start.size(0))
+
+ nb_fish = torch.randint(max_nb_obj, (1,)).item() + 1
+ for c in (
+ (torch.randperm(nb_fish_tokens) + first_fish_token)[:nb_fish].sort().values
+ ):
+ i, j = (
+ torch.randint(height - 2, (1,))[0] + 1,
+ torch.randint(width - 2, (1,))[0] + 1,
+ )
+ vm = torch.randint(4, (1,))[0]
+ vi, vj = (vm // 2) * (2 * (vm % 2) - 1), (1 - vm // 2) * (2 * (vm % 2) - 1)
+
+ f_start[i, j] = c
+ f_start[i - vi, j - vj] = c
+ f_start[i + vj, j - vi] = c
+ f_start[i - vj, j + vi] = c
+
+ for l in range(nb_iterations):
+ i += vi
+ j += vj
+ if i < 0 or i >= height or j < 0 or j >= width:
+ i -= vi
+ j -= vj
+ vi, vj = -vi, -vj
+ i += vi
+ j += vj
+
+ f_end[i, j] = c
+ f_end[i - vi, j - vj] = c
+ f_end[i + vj, j - vi] = c
+ f_end[i - vj, j + vi] = c
+
+ pairs.append((f_start, f_end))
+
+ result = []
+ for p in pairs:
+ if torch.rand(1) < 0.5:
+ result.append(
+ torch.cat(
+ [p[0].flatten(), torch.tensor([token_forward]), p[1].flatten()],
+ dim=0,
+ )[None, :]
+ )
+ else:
+ result.append(
+ torch.cat(
+ [p[1].flatten(), torch.tensor([token_backward]), p[0].flatten()],
+ dim=0,
+ )[None, :]
+ )
+
+ return torch.cat(result, dim=0)
+
+
+def sample2img(seq, height, width, upscale=15):
+ f_first = seq[:, : height * width].reshape(-1, height, width)
+ f_second = seq[:, height * width + 1 :].reshape(-1, height, width)
+ direction = seq[:, height * width]
+
+ def mosaic(x, upscale):
+ x = x.reshape(-1, height, width)
+ m = torch.logical_and(x >= 0, x < first_fish_token + nb_fish_tokens).long()
+ x = colors[x * m].permute(0, 3, 1, 2)
+ s = x.shape
+ x = x[:, :, :, None, :, None].expand(-1, -1, -1, upscale, -1, upscale)
+ x = x.reshape(s[0], s[1], s[2] * upscale, s[3] * upscale)
+
+ for n in range(m.size(0)):
+ for i in range(m.size(1)):
+ for j in range(m.size(2)):
+ if m[n, i, j] == 0:
+ for k in range(2, upscale - 2):
+ x[n, :, i * upscale + k, j * upscale + k] = 0
+ x[n, :, i * upscale + upscale - 1 - k, j * upscale + k] = 0
+
+ return x
+
+ direction_symbol = torch.full((direction.size(0), height * upscale, upscale), 0)
+ direction_symbol = colors[direction_symbol].permute(0, 3, 1, 2)
+ separator = torch.full((direction.size(0), 3, height * upscale, 1), 0)
+
+ for n in range(direction_symbol.size(0)):
+ if direction[n] == token_forward:
+ for k in range(upscale):
+ direction_symbol[
+ n,
+ :,
+ (height * upscale) // 2 - upscale // 2 + k,
+ 3 + abs(k - upscale // 2),
+ ] = 0
+ elif direction[n] == token_backward:
+ for k in range(upscale):
+ direction_symbol[
+ n,
+ :,
+ (height * upscale) // 2 - upscale // 2 + k,
+ 3 + upscale // 2 - abs(k - upscale // 2),
+ ] = 0
+ else:
+ for k in range(2, upscale - 2):
+ direction_symbol[
+ n, :, (height * upscale) // 2 - upscale // 2 + k, k
+ ] = 0
+ direction_symbol[
+ n, :, (height * upscale) // 2 - upscale // 2 + k, upscale - 1 - k
+ ] = 0
+
+ return torch.cat(
+ [
+ mosaic(f_first, upscale),
+ separator,
+ direction_symbol,
+ separator,
+ mosaic(f_second, upscale),
+ ],
+ dim=3,
+ )
+
- while True:
-
- frames =[]
-
- scene = random_scene()
- xh, yh = tuple(x.item() for x in torch.rand(2))
-
- frames.append(scene2tensor(xh, yh, scene))
-
- actions = torch.randint(len(effects), (nb_steps,))
- change = False
-
- for a in actions:
- g, dx, dy = effects[a]
- if g:
- for b in scene:
- if b.x <= xh and b.x + b.w >= xh and b.y <= yh and b.y + b.h >= yh:
- x, y = b.x, b.y
- b.x += dx
- b.y += dy
- if (
- b.x < 0
- or b.y < 0
- or b.x + b.w > 1
- or b.y + b.h > 1
- or b.collision(scene)
- ):
- b.x, b.y = x, y
- else:
- xh += dx
- yh += dy
- change = True
- else:
- x, y = xh, yh
- xh += dx
- yh += dy
- if xh < 0 or xh > 1 or yh < 0 or yh > 1:
- xh, yh = x, y
-
- if all_frames:
- frames.append(scene2tensor(xh, yh, scene))
-
- if not all_frames:
- frames.append(scene2tensor(xh, yh, scene))
-
- if change:
- break
-
- return frames, actions
+def seq2str(seq):
+ result = []
+ for s in seq:
+ result.append("".join([token2char[v] for v in s]))
+ return result
+######################################################################
+
if __name__ == "__main__":
- frames, actions = sequence(nb_steps=31,all_frames=True)
- frames = torch.cat(frames,0)
- print(f"{frames.size()=}")
- torchvision.utils.save_image(frames, "seq.png", nrow=8)
+ import time
+
+ height, width = 6, 8
+ start_time = time.perf_counter()
+ seq = generate(nb=90, height=height, width=width, max_nb_obj=3)
+ delay = time.perf_counter() - start_time
+ print(f"{seq.size(0)/delay:02f} samples/s")
+
+ print(seq2str(seq[:4]))
+
+ # m = (torch.rand(seq.size()) < 0.05).long()
+ # seq = (1 - m) * seq + m * 23
+
+ img = sample2img(seq, height, width)
+ print(img.size())
+
+ torchvision.utils.save_image(
+ img.float() / 255.0, "/tmp/world.png", nrow=6, padding=4
+ )