captcha-solver/solve_captcha.py

import tensorflow as tf
from tensorflow.keras.models import load_model
from imutils import paths
import numpy as np
import cv2
import pickle
from tratar_captcha import tratar_base64
import base64

tamanho_imagem = (35, 50)
modelo = load_model("modelo_treinado.keras")

# Posições fixas para o corte de cada caractere
posicoes_corte = [(6, 40), (56, 90), (111, 144), (161, 197), (214, 250), (265, 297)]

def quebrar_captcha_base64(base64_string):
    if base64_string.startswith('data:image/png;base64,'):
        base64_string = base64_string.replace('data:image/png;base64,', '')
    
    with open("rotulos_modelo.dat", "rb") as arquivo_tradutor:
        lb = pickle.load(arquivo_tradutor)
    
    imagem_tratada = tratar_base64(base64_string)

    # Aplicar um limiar para binarizar a imagem
    _, imagem_bin = cv2.threshold(imagem_tratada, 128, 255, cv2.THRESH_BINARY_INV)

    imagem_final = cv2.merge([imagem_bin] * 3)  # Convertendo de volta para RGB para desenhar os retângulos coloridos

    # Processar cada região de corte definida
    previsao = []
    for x1, x2 in posicoes_corte:
        imagem_caractere = imagem_bin[20:-15, x1:x2]  # Cortar a região do caractere
        imagem_caractere = cv2.bitwise_not(imagem_caractere)
        
        imagem_caractere = cv2.resize(imagem_caractere, tamanho_imagem)

        imagem_caractere = np.expand_dims(imagem_caractere, axis=2)
        imagem_caractere = np.expand_dims(imagem_caractere, axis=0)

        letra_prevista = modelo.predict(imagem_caractere)
        letra_prevista = lb.inverse_transform(letra_prevista)[0]
        previsao.append(letra_prevista)

    texto_previsao = "".join(previsao)

    print(texto_previsao)

    return texto_previsao


# Exemplo de uso
if __name__ == "__main__":
    # Aqui você deve substituir por uma string base64 real
    base64_string = "data:image/png;base64, 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"
    
    quebrar_captcha_base64(base64_string)