app.py

import gradio as gr
import tensorflow as tf
import numpy as np
import matplotlib.image as mpimg
from tensorflow.keras import  models


model = models.load_model('alfabot.keras')
UPPERCASE_ALFABET = ["A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z"]

def predict(img):
    image = mpimg.imread(img)
    image = np.expand_dims(image, axis=0)  
    image_tensor = tf.constant(image, dtype=tf.float32)
    result = model.predict(image_tensor)[0]
    return UPPERCASE_ALFABET[np.argmax(result)]


gr.Interface(
    fn=predict,
    inputs=gr.inputs.Image(shape=(28, 28)),
    outputs=gr.outputs.Label(num_top_classes=3),
    title="Identificador de letras manuscritas",
    description="Esse modelo tem a capacidade de identificar qual é a letra manuscrita.",
    examples=["A.jpg","B.jpg","C.jpg","D.jpg","E.jpg"],
).launch()