import os
import numpy as np
from PIL import Image, ImageDraw
from tensorflow import keras


characters = "abcdefghijklmnopqrstuvwxyz"
num_classes = len(characters)

ascii_art_folder = "dataset"
image_folder = "images"

for character in characters:
    ascii_file = os.path.join(ascii_art_folder, character + ".txt")
    image_file = os.path.join(image_folder, character + ".png")

    with open(ascii_file, "r") as file:
        ascii_art = file.read()

    image = Image.new("RGB", (100, 100), color="white")
    image_draw = ImageDraw.Draw(image)
    image_draw.text((10, 10), ascii_art, fill="black")
    image.save(image_file)

samples = []
labels = []
for character in characters:
    image_file = os.path.join(image_folder, character + ".png")

    image = Image.open(image_file).convert("L")

    image = image.resize((64, 64))

    image_array = np.array(image)

    image_array = image_array / 255.0

    samples.append(image_array)
    labels.append(characters.index(character))

samples = np.array(samples)
labels = np.array(labels)

# Split the dataset into training and testing
split_ratio = 0.8
split_index = int(len(samples) * split_ratio)

train_samples = samples[:split_index]
train_labels = labels[:split_index]
test_samples = samples[split_index:]
test_labels = labels[split_index:]

# Define the model architecture
model = keras.models.Sequential([
    keras.layers.Conv2D(32, (3, 3), activation="relu", input_shape=(64, 64, 1)),
    keras.layers.MaxPooling2D((2, 2)),
    keras.layers.Flatten(),
    keras.layers.Dense(64, activation="relu"),
    keras.layers.Dense(num_classes, activation="softmax")
])

# Compile the model
model.compile(optimizer="adam", loss="sparse_categorical_crossentropy", metrics=["accuracy"])

# Train the model
model.fit(train_samples, train_labels, epochs=100, batch_size=32, validation_split=0.2)

# Evaluate the model
test_loss, test_acc = model.evaluate(test_samples, test_labels)
print(f"Test accuracy: {test_acc}")

# Convert input ASCII art to image and predict the letter
input_ascii_art = """
  _   _
 | | | |
 | |_| |
  \__, |
  |___/
"""
input_image = Image.new("RGB", (100, 100), color="white")
input_image_draw = ImageDraw.Draw(input_image)
input_image_draw.text((10, 10), input_ascii_art, fill="black")
input_image = input_image.convert("L")
input_image = input_image.resize((64, 64))
input_image_array = np.array(input_image) / 255.0
input_image_array = np.expand_dims(input_image_array, axis=0)
input_image_array = np.expand_dims(input_image_array, axis=-1)

prediction = model.predict(input_image_array)
predicted_index = np.argmax(prediction)
predicted_character = characters[predicted_index]

print(f"Predicted character: {predicted_character}")