Words that have rotational symetry are called "ambigrams"
I find most of these online are using tricky font magic where multiple letters are vauge and merged together. But in english there are lots of letters that allready have rotational symetry, such as b and q and d and p.
Lets see if this is actually a bad idea? See below how I search and find some phrases that work.
# Imports
import os
import cv2
import numpy as np
import matplotlib.pyplot as plt# Letters in english that I was able to form through rotational symmetry
# A B D E F H I L M N O X Z S G J P Q T U W Y
# Letters i failed to do so, you evil characters
# C V K R# Mapping of those letters
matches = [
('x','x'),
('o','o'),
('z','z'),
('l','l'),
('m','w'),
('n','u'),
('d','p'),
('b','q'),
('e','a'),
('h','y'),
('i','t'),
('f','j'),
('h','h'),
('s','s'),
('g','g'),
]# Load images
font_raw = dict()
files = os.listdir('../font')
for f in files:
loaded_img = cv2.imread('../font/' + f)
name = f.split('.')[0].lower()
font_raw[name] = loaded_img# Make fonts cleanly sized
def crop ( img ):
def cut ( img ):
size = img.shape[0] // 2
for i in range(2,9):
cut = img[:size, :]
if np.min(cut) < 240:
size -= ( img.shape[0] // (2**i))
else:
size += ( img.shape[0] // (2**i))
return img[size:,:]
img = cut(img)
img = cv2.flip(img, 0)
img = cut(img)
img = cv2.transpose(img)
img = cut(img)
img = cv2.flip(img, 0)
img = cut(img)
img = cv2.transpose(img)
img = cv2.flip(img, 0)
img = cv2.flip(img, 1)
# Pad a bit
img = cv2.copyMakeBorder(img, 10, 10, 10, 10, cv2.BORDER_CONSTANT, value=[255,255,255])
return imgplt.imshow(crop(font_raw['b']))<matplotlib.image.AxesImage at 0x42df62c10>
# Crop all the letter and assign pairs
font_full = dict()
# Do crop
for k,v in font_raw.items():
font_full[k] = crop(v)
# Do matching
for (a,b) in matches:
if a not in font_full:
print('Missing letter: ' + a)
else:
_ = cv2.rotate(font_full[a], cv2.ROTATE_90_CLOCKWISE)
_ = cv2.rotate(_, cv2.ROTATE_90_CLOCKWISE)
font_full[b] = _def merge (images):
size = 0
cursor = 0
for image in images:
size += image.shape[1]
img = np.zeros((200, size, 3), np.uint8)
img[:,:] = 255
for i,image in enumerate(images):
img[:image.shape[0], cursor:(cursor+image.shape[1])] = image
cursor += image.shape[1]
return img# Loading all our letters into a single image
print("Not the prettiest, never claimed I was an artist")
letters = sorted(list(font_full.keys()))
my_alphabet = merge([ font_full[v] for v in letters])
plt.figure(figsize = (20,2))
plt.imshow(my_alphabet)Not the prettiest, never claimed I was an artist
<matplotlib.image.AxesImage at 0x42e022040>
# Load words from file
with open('words.txt') as f:
words = set(f.read().splitlines())match_dict = {}
for match in matches:
match_dict[match[0]] = match[1]
match_dict[match[1]] = match[0]def try_flip (word) :
result = list(word)
for i, letter in enumerate(word):
if letter in match_dict:
result[len(word)-i-1] = match_dict[letter]
else:
return False
return ''.join(result)def valid_phrase (phrase) :
for word in phrase.split(' '):
if word not in words:
return False
return True# What words are valid if we flip the image?
results = []
for word in words:
flipped = try_flip(word)
if flipped and valid_phrase(flipped):
results.append((word, flipped))
# Sort by length
results.sort(key=lambda x: -len(x[0]))print("These words are valid if you flip the image")
top_single_words = results[0:10]
top_single_words[('play', 'held'), ('age', 'age'), ('it', 'it'), ('am', 'we'), ('we', 'am')]
def render (word):
size = 0
cursor = 0
for letter in word:
size += font_full[letter].shape[1]
img = np.zeros((200, size, 3), np.uint8)
img[:,:] = 255
for i, letter in enumerate(word):
img[:font_full[letter].shape[0], cursor:(cursor+font_full[letter].shape[1])] = font_full[letter]
cursor += font_full[letter].shape[1]
return imgdef render_both ( word ) :
result = render( word )
_ = cv2.rotate(result, cv2.ROTATE_90_CLOCKWISE)
flipped = cv2.rotate(_, cv2.ROTATE_90_CLOCKWISE)
plt.imshow(result)
plt.show()
plt.imshow(flipped)
plt.show()# LOOK! WE FOUND ONE!
render_both("play")
all_phrases = set()
all_flippable_words = set()
# Get flippable
for word in words:
flipped = try_flip(word)
if len(word) < 6 and flipped:
all_flippable_words.add(word)
print("Found {} flippable words".format(len(all_flippable_words)))
# Add Phrases
for word in all_flippable_words:
all_phrases.add(word)
for word2 in all_flippable_words:
all_phrases.add(word + word2)
for word3 in all_flippable_words:
all_phrases.add(word + word2 + word3)
print ("Found {} phrases".format(len(all_phrases)))Found 399 flippable words
Found 63567955 phrases
# Search for reflections
results = []
total = len(all_phrases)
for i, phrases in enumerate(all_phrases):
flipped = try_flip(phrases)
if (i % 1000) == 0:
print(f"Checking {i}/{total}: {phrases}", end='\r')
if flipped and flipped in all_phrases:
results.append((phrases, flipped))Checking 63567000/63567955: eatmoneywillyln
## Sort by length
results.sort(key=lambda x: -len(x[0]))
results[:10][('playplayplay', 'heldheldheld'),
('putmindplay', 'heldputwind'),
('putwindplay', 'heldputmind'),
('playputwind', 'putmindheld'),
('playputmind', 'putwindheld'),
('putwithplay', 'heldhitmind'),
('playageplay', 'heldageheld'),
('hitmindplay', 'heldputwith'),
('playhitmind', 'putwithheld'),
('ageplayplay', 'heldheldage')]
# And some of these almost make sense haha
render_both('putwindplay')
import random
for i in range(10):
(a,_) = random.choice(results)
print("Ambigram:", a,_)
render_both(a)
plt.show()Ambigram: golaysit itshelog
Ambigram: iseaseit itaseast
Ambigram: golayage agehelog
Ambigram: ithit ithit
Ambigram: ohmeplay heldawho
Ambigram: mesaysay heshesaw
Ambigram: mesatpage agediesaw
Ambigram: playitsay hesitheld
Ambigram: itheadoh hopeahit
Ambigram: agegopay hedogage
