I read a post about visually symmetric words a while back and decided to write a program to automatically find them.
The code can be found on on github at mark-veres/vis-sym.

I began by finding all visually symmetrical letters and ended up with this list:

symmetric = ['i', 'l', 'm', 'o', 'u', 'v', 'w', 'x', 'y']

I created a list that contained pairs of mirrored letters.

mirrors = [('p', 'q'), ('d', 'b'), ('q', 'p'), ('b', 'd')]
for s in symmetric:
    mirrors.append((s, s))

The is_symmetric function takes a word, splits it in two parts (and optionally the character in the middle of the word) and checks if the letters of the words are mirrors or visually symmetric.

For a word to be visually symmetric:

• middle must be symmetric
• every letter in the first half must be equal to or a mirror of his opposite

def is_symmetric(word):
    if len(word)%2 == 0:
        first_half = word[:int(len(word)/2)]
        second_half = word[int(len(word)/2):] [::-1]
    else:
        first_half = word[:int((len(word)-1)/2)]
        second_half = word[int((len(word)+1)/2):] [::-1]
        middle = word[int((len(word)-1)/2)]

        if middle not in symmetric:
            return False
    
    for i, char in enumerate(first_half):
        opposite = second_half[i]
        if (char, opposite) not in mirrors:
            return False

    return True

The [::-1] expression is used to reverse the second_half string, to make the comparing of characters easier.

The final just takes words from a dictionary and checks if they’re visually symmetric with the is_symmetric function.