!ls /content/dataset/colectivos | wc -l
!ls /content/dataset/ventiladores | wc -l

https://th.bing.com/th/id/OIP.u_k7dnKroJnwvlqI8SMjhQHaFj?w=960&h=720&rs=1&pid=ImgDetMain325
407

# IMPORTACION DE LIBRERIAS
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import numpy as np

#Crear el dataset generador
datagen = ImageDataGenerator(
    rescale=1. / 255,
    rotation_range = 30,
    width_shift_range=0.15,
    height_shift_range = 0.15,
    shear_range = 5,
    zoom_range = [0.7, 1.3],
    validation_split = 0.2
)

data_gen_entrenamiento = datagen.flow_from_directory("/content/dataset",
                                                     target_size=(224,224),
                                                     batch_size=32, shuffle=True,
                                                     subset="training")

data_gen_pruebas = datagen.flow_from_directory("/content/dataset",
                                                     target_size=(224,224),
                                                     batch_size=32, shuffle=True,
                                                     subset="validation")

Found 586 images belonging to 3 classes.
Found 146 images belonging to 3 classes.

import matplotlib.pyplot as plt

for imagenes, etiquetas in data_gen_entrenamiento:
  for i in range(10):
    plt.subplot(2, 5, i+1)
    plt.imshow(imagenes[i])

  break
plt.show()

import tensorflow as tf
import tensorflow_hub as hub

url ="https://www.tfhub.dev/google/tf2-preview/mobilenet_v2/feature_vector/4"
mobilenetv2 = hub.KerasLayer(url, input_shape=(224, 224, 3))

# congelamiento de las capas para evitar overfitting
mobilenetv2.trainable = False

modelo = tf.keras.Sequential([
    mobilenetv2,
    tf.keras.layers.Dense(3, activation="softmax")
])

modelo.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 keras_layer (KerasLayer)    (None, 1280)              2257984   
                                                                 
 dense (Dense)               (None, 3)                 3843      
                                                                 
=================================================================
Total params: 2261827 (8.63 MB)
Trainable params: 3843 (15.01 KB)
Non-trainable params: 2257984 (8.61 MB)
_________________________________________________________________

modelo.compile(
    optimizer="adam",
    loss="categorical_crossentropy",
    metrics=["accuracy"]
)

EPOCAS = 10
historial = modelo.fit(
    data_gen_entrenamiento, epochs=EPOCAS, batch_size=32,
    validation_data=data_gen_pruebas
)

Epoch 1/10

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:996: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  warnings.warn(

 9/19 [=============>................] - ETA: 4s - loss: 0.6891 - accuracy: 0.6562

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:996: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  warnings.warn(

19/19 [==============================] - ETA: 0s - loss: 0.4397 - accuracy: 0.7969

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:996: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  warnings.warn(


19/19 [==============================] - 22s 760ms/step - loss: 0.4397 - accuracy: 0.7969 - val_loss: 0.0865 - val_accuracy: 0.9863
Epoch 2/10
 2/19 [==>...........................] - ETA: 9s - loss: 0.0814 - accuracy: 0.9762

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:996: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  warnings.warn(

18/19 [===========================>..] - ETA: 0s - loss: 0.1021 - accuracy: 0.9621

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:996: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  warnings.warn(

from PIL import Image
import cv2

def categorizar(ruta):
  img = Image.open(ruta)
  img = img.convert("RGB")
  img = np.array(img).astype(float)/255

  img = cv2.resize(img, (224, 224))
  prediccion = modelo.predict(img.reshape(-1, 224, 224, 3))
  return np.argmax(prediccion[0], axis=-1)

# graficar precision
acc= historial.history['accuracy']
val_acc = historial.history['val_accuracy']

loss= historial.history['loss']
val_loss = ['val_loss']

rango_epocas=range(10)

plt.figure(figsize=(8,8))
plt.subplot(1,2,1)
plt.plot(rango_epocas , acc , label = 'Precisión Entrenamiento')
plt.plot(rango_epocas, val_acc , label='Precisión Pruebas')
plt.legend(loc='upper right')
plt.title('Precisión del entrenamiento y Pruebas')

plt.subplot(1,2,2)
plt.plot(rango_epocas , loss , label='Perdida de Entrenamiento')
# plt.plot(rango_epocas, val_loss, label='Pérdida de Pruebas')
plt.legend(loc='upper right')
plt.title('Precisión del entrenamiento y Pruebas')

plt.show()

import requests
from io import BytesIO
import cv2

def categorizacion(url):
  respuesta = requests.get(url)
  img = Image.open(BytesIO(respuesta.content))
  img = np.array(img).astype(float)/255

  img = cv2.resize(img , (224,224))
  prediccion= modelo.predict(img.reshape(-1 , 224 , 224 , 3))
  return np.argmax(prediccion[0], axis=-1)

while True:
  try:
    url = input("Introduce aqui el link de la Imagen : ")
    prediccion= categorizacion(url)
    if prediccion == 1:
      print("La imagen Corresponde a un Colectivo")
    elif prediccion == 2:
      print("La Imagen Corresponde a un Ventilador")

  except :
     print("No se pudo encontrar la direccion de la Imagen")

No se pudo encontrar la direccion de la Imagen
Introduce aqui el link de la Imagen : https://th.bing.com/th/id/OIP.q4uy1CnD7dD9jt-rVx9aHgHaCv?w=332&h=129&c=7&r=0&o=5&pid=1.7
1/1 [==============================] - 0s 33ms/step
La imagen Corresponde a un Colectivo
Introduce aqui el link de la Imagen : https://th.bing.com/th/id/OIP.TpRiILM9_eQL2Ii9LGZ-yAHaGY?w=251&h=216&c=7&r=0&o=5&pid=1.7
1/1 [==============================] - 0s 108ms/step
La Imagen Corresponde a un Ventilador