Axwell15/AN_proyecto

1import gradio as gr2import csv3import os4import numpy as np5from sklearn.datasets import load_iris6from sklearn.preprocessing import StandardScaler7 8 9# ---------CONFIGURACIÓN DE HISTORIAL----------10HISTORY_FILE = "historial.csv"11 12def load_history():13    if os.path.exists(HISTORY_FILE):14        with open(HISTORY_FILE, mode="r", newline="") as f:15            reader = csv.reader(f)16            #Convirtiendo a float 17            return[18                [float(row[0]), float(row[1]), float(row[2]), float(row[3]), row[4]]19                for row in reader20            ]21    return []22 23def save_history_row(row):24    with open(HISTORY_FILE, mode="a", newline="") as f:25        writer = csv.writer(f)26        writer.writerow(row)27 28def clear_history_file():29    if os.path.exists(HISTORY_FILE):30        os.remove(HISTORY_FILE)31 32 33# Cargar el conjunto de datos Iris34iris = load_iris()35X_iris = iris.data36y_iris = iris.target37 38# --------- Custom Logistic Regression (One-vs-Rest, Gradient Descent) ---------39def sigmoid(z):40    return 1 / (1 + np.exp(-z))41 42def gradient_descent(X, y, lr=0.1, epochs=1000):43    m, n = X.shape44    X = np.hstack((np.ones((m,1)), X))45    weights = np.zeros(n + 1)46    for _ in range(epochs):47        z = np.dot(X, weights)48        h = sigmoid(z)49        grad = np.dot(X.T, (h - y)) / m50        weights -= lr * grad51    return weights52 53def predict_custom(X, weights_all):54    X = np.hstack((np.ones((X.shape[0],1)), X))55    preds = np.array([sigmoid(np.dot(X, w)) for w in weights_all]).T56    return np.argmax(preds, axis=1)57 58# Escalar los datos para el modelo personalizado59scaler_custom = StandardScaler()60X_iris_scaled = scaler_custom.fit_transform(X_iris)61 62# Entrenar modelo personalizado One-vs-Rest63weights_all_custom = []64for class_label in np.unique(y_iris):65    y_binary = (y_iris == class_label).astype(int)66    weights = gradient_descent(X_iris_scaled, y_binary, lr=0.1, epochs=1000)67    weights_all_custom.append(weights)68 69 70# ------------------- Función de predicción -------------------71def predict(sepal_length, sepal_width, petal_length, petal_width):72    if all(float(val) == 0.0 for val in [sepal_length, sepal_width, petal_length, petal_width]):73        return "⚠️ Por favor ingresa valores válidos (mayores a cero).", load_history(), None74    input_data = np.array([[sepal_length, sepal_width, petal_length, petal_width]])75    input_scaled = scaler_custom.transform(input_data)76    pred_idx = predict_custom(input_scaled, weights_all_custom)[0]77    class_name = iris.target_names[pred_idx]78    result = f"Predicción (Regresión Logística Custom GD): {class_name}"79 80    image_path = f"./imagenes/{class_name}.jpg"81    row = [sepal_length, sepal_width, petal_length, petal_width, class_name]82    83    history = load_history()84    history.append(row)85    save_history_row(row)86 87    return result, history, image_path88 89 90 91# ------------------- Función para limpiar las entradas -------------------92def clear_inputs():93    clear_history_file()94    return 0.0, 0.0, 0.0, 0.0, [], None95 96# ------------------- Interfaz de Gradio -------------------97def iris_interface():98    history = load_history()99    with gr.Blocks() as demo:100        gr.Markdown("## Clasificación Iris con Regresión Logística (Descenso de Gradiente Custom)")101    with gr.Row():102        with gr.Column(scale=2):103            with gr.Row():104                sepal_length = gr.Number(label="Sepal Length (cm)", value=0.0)105                sepal_width = gr.Number(label="Sepal Width (cm)", value=0.0)106                petal_length = gr.Number(label="Petal Length (cm)", value=0.0)107                petal_width = gr.Number(label="Petal Width (cm)", value=0.0)108                output = gr.Textbox(label="Output")109        with gr.Column(scale=4):110            with gr.Row():111                with gr.Column(scale=1):112                    image_output = gr.Image(113                                label="Imagen de la Flor",114                                type="filepath",115                                interactive=False,116                                height=267,117                                width=300,118                          )119                clear_button = gr.Button("Clear")120                submit_button = gr.Button("Submit")121                        122 123    history_output = gr.Dataframe(124        headers=["Sepal Length", "Sepal Width", "Petal Length", "Petal Width", "Predicción"],125        label="Historial de Predicciones",126        interactive=False,127        wrap=True,128        value=history 129    )130    131    submit_button.click(132        predict,133        inputs=[sepal_length, sepal_width, petal_length, petal_width],134        outputs=[output, history_output, image_output]135    )136    137    clear_button.click(138        clear_inputs,139        inputs=[],140        outputs=[sepal_length, sepal_width, petal_length, petal_width, history_output, image_output]141    )142 143 144# Lanzar la aplicación145    return demo146