App.py

app.py

@@ -1,127 +1,40 @@
-# -*- coding: utf-8 -*-
-"""ProyectoAvanzado1.ipynb
-
-Automatically generated by Colab.
-
-Original file is located at
-    https://colab.research.google.com/drive/1R5pOFRcOn9faAiaFmIL6GcuWweWYGswX
-"""
-
-import os
-import pickle
+import streamlit as st
 import numpy as np
-from sklearn.model_selection import train_test_split
-from tensorflow.keras.utils import to_categorical
+import pandas as pd
 import matplotlib.pyplot as plt
 import seaborn as sns
-import pandas as pd
-from sklearn import preprocessing
-import streamlit as st
-import tensorflow as tf
-import numpy as np
 import cv2
+import torch
+from transformers import AutoTokenizer, AutoModelForImageClassification
+from huggingface_hub import hf_hub_download
 from PIL import Image
-import matplotlib.pyplot as plt
-import seaborn as sns
-import tensorflow as tf
-from tensorflow.keras.models import Sequential
-from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout
+import pickle
 
-# Función para cargar datos desde archivos .p
-def load_pickle(file_path):
-    with open(file_path, 'rb') as file:
-        data = pickle.load(file)
-    return data
+# Cargar dataset
+dataset_repo = "VicGerardoPR/Traffic_Sign_Classifier"
+dataset_file = "train.p"
 
-# Cargar datos de entrenamiento, validación y prueba
-train_data = load_pickle(os.path.join("/Users/victorg/Documents/DataScience/ProjectoAvanzado/CSV/train.p"))
-val_data = load_pickle(os.path.join("/Users/victorg/Documents/DataScience/ProjectoAvanzado/CSV/valid.p"))
-test_data = load_pickle(os.path.join("/Users/victorg/Documents/DataScience/ProjectoAvanzado/CSV/test.p"))
+dataset_path = hf_hub_download(repo_id=dataset_repo, filename=dataset_file)
+with open(dataset_path, 'rb') as file:
+    train_data = pickle.load(file)
 
-# Separar imágenes y etiquetas
 train_images, train_labels = train_data['features'], train_data['labels']
-val_images, val_labels = val_data['features'], val_data['labels']
-test_images, test_labels = test_data['features'], test_data['labels']
 
 # Preprocesar datos
 def preprocess_data(images, labels):
     images = images.astype('float32') / 255.0
-    labels = to_categorical(labels, num_classes=43)
+    labels = pd.Series(labels).astype('category').cat.codes.to_numpy()
     return images, labels
 
 train_images, train_labels = preprocess_data(train_images, train_labels)
-val_images, val_labels = preprocess_data(val_images, val_labels)
-test_images, test_labels = preprocess_data(test_images, test_labels)
-
-
-# Convertir etiquetas a DataFrame para visualización
-train_df = pd.DataFrame({"label": train_labels.argmax(axis=1)})
-
-# Conteo de imágenes por clase
-plt.figure(figsize=(10, 6))
-sns.countplot(x=train_df['label'])
-plt.title("Conteo de Imágenes por Clase")
-plt.xlabel("Clase")
-plt.ylabel("Conteo")
-plt.show()
-
-# Visualización 2: Ejemplos de imágenes por clase
-fig, axes = plt.subplots(5, 5, figsize=(15, 15))
-axes = axes.ravel()
-for i in range(25):
-    axes[i].imshow(train_images[i])
-    axes[i].set_title(f"Clase: {train_labels[i].argmax()}")
-    axes[i].axis('off')
-plt.subplots_adjust(hspace=0.5)
-plt.show()
-
-# Visualización 3: Muestra aleatoria de imágenes
-fig, axes = plt.subplots(3, 3, figsize=(12, 12))
-axes = axes.ravel()
-for i in range(9):
-    idx = np.random.randint(0, len(train_images))
-    axes[i].imshow(train_images[idx])
-    axes[i].set_title(f"Clase: {train_labels[idx].argmax()}")
-    axes[i].axis('off')
-plt.subplots_adjust(hspace=0.5)
-plt.show()
-
-
-
-# Definir la arquitectura de la CNN
-model = Sequential([
-    Conv2D(32, (3, 3), activation="relu", input_shape=(32, 32, 3)),
-    MaxPooling2D((2, 2)),
-    Conv2D(64, (3, 3), activation="relu"),
-    MaxPooling2D((2, 2)),
-    Conv2D(128, (3, 3), activation="relu"),
-    MaxPooling2D((2, 2)),
-    Flatten(),
-    Dense(215, activation="relu"),
-    Dropout(0.5),
-    Dense(43, activation="softmax")
-])
-
-# Compilar el modelo
-model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])
-
-# Verificar si el modelo ya está guardado
-model_path = "traffic_sign_classifier.h5"
-if os.path.exists(model_path):
-    model = tf.keras.models.load_model(model_path)
-else:
-    model.fit(train_images, train_labels, epochs=10, validation_data=(val_images, val_labels))
-    model.save(model_path)
 
-# Evaluar el modelo en datos de prueba
-test_loss, test_acc = model.evaluate(test_images, test_labels)
-print(f"Test Accuracy: {test_acc:.2f}")
+# Cargar modelo
+model_repo = "VicGerardoPR/Traffic_Sign_Classifier"
+model_file = "traffic_sign_classifier.h5"
 
-
-
-
-# Cargar el modelo
-model = tf.keras.models.load_model("traffic_sign_classifier.h5")
+model_path = hf_hub_download(repo_id=model_repo, filename=model_file)
+model = torch.load(model_path, map_location=torch.device('cpu'))
+model.eval()
 
 # Diccionario de clases de señales de tráfico
 classes = {
@@ -170,14 +83,17 @@ classes = {
     42: 'End of no passing by vehicles over 3.5 metric tons'
 }
 
-# Función para predecir la clase de una imagen
 def predict(image):
     image = np.array(image)
     image = cv2.resize(image, (32, 32))
     image = image / 255.0
     image = np.expand_dims(image, axis=0)
-    predictions = model.predict(image)
-    class_idx = np.argmax(predictions)
+    image = np.transpose(image, (0, 3, 1, 2))  # Reordenar dimensiones para PyTorch
+    image = torch.tensor(image).float()
+
+    with torch.no_grad():
+        predictions = model(image)
+        class_idx = predictions.argmax().item()
     return classes[class_idx]
 
 # Título y descripción de la aplicación
@@ -191,29 +107,10 @@ axes = axes.ravel()
 for i in range(10):
     idx = np.random.randint(0, len(train_images))
     axes[i].imshow(train_images[idx])
-    axes[i].set_title(f"Clase: {train_labels[idx].argmax()}")
-    axes[i].axis('off')
-st.pyplot(fig)
-# Mostrar ejemplos de imágenes del conjunto de datos
-
-fig, axes = plt.subplots(2, 5, figsize=(15, 6))
-axes = axes.ravel()
-for i in range(10):
-    idx = np.random.randint(0, len(train_images))
-    axes[i].imshow(train_images[idx])
-    axes[i].set_title(f"Clase: {train_labels[idx].argmax()}")
+    axes[i].set_title(f"Clase: {train_labels[idx]}")
     axes[i].axis('off')
 st.pyplot(fig)
 
-# Mostrar algunas visualizaciones de EDA
-st.header("Visualizaciones de EDA")
-fig, ax = plt.subplots(figsize=(10, 6))
-sns.countplot(x='label', data=train_df, ax=ax)
-ax.set_title("Conteo de Imágenes por Clase")
-ax.set_xlabel("Clase")
-ax.set_ylabel("Conteo")
-st.pyplot(fig)
-
 # Permitir al usuario cargar una imagen
 st.header("Carga tu Propia Imagen de Señal de Tráfico")
 uploaded_file = st.file_uploader("Elige una imagen...", type=["jpg", "jpeg", "png"])
@@ -223,8 +120,4 @@ if uploaded_file is not None:
     st.write("")
     st.write("Clasificando...")
     label = predict(image)
-    st.write(f"Esta señal es: {label}")
-
-# Mostrar métricas del modelo
-st.header("Métricas del Modelo")
-st.write(f"Exactitud del conjunto de prueba: {test_acc:.2f}")
+    st.write(f"Esta señal es: {label}")