Модель CNN не может классифицировать изображение фруктов даже при высокой точности обучения

Question

Я использую набор данных kaggle fruit 360 и хочу спрогнозировать 10-15 фруктов, и модель мне нравится, точность тренировочного тестирования также хорошая. но когда в конце я загружаю какое-то изображение и кормлю его для проверки модели, он полностью распознает его как другой фрукт, кроме яблочного. Может быть, это в конце модели, где я проверяю соответствие указанного класса c. Т.е. последний кусок, куда я загружаю изображение. Я пробовал модель VGG16, и даже у нее такая же проблема с более высокой точностью. Любая помощь будет принята с благодарностью, так как я застрял на ней последние два дня.

это мой код:

"""Fruits.ipynb
Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1AiGMVBshT9w-5T3fzzi3jyrJr0CVGv6A
"""

from google.colab import files

!pip install -q kaggle

uploaded = files.upload()

!mkdir -p ~/.kaggle
!cp kaggle.json ~/.kaggle/
!ls ~/.kaggle
!chmod 600 /root/.kaggle/kaggle.json

!kaggle kernels list — user hinaaslam1 — sort-by dateRun

!kaggle datasets download -d moltean/fruits 

#https://www.kaggle.com/moltean/fruits see the dataset at this link.

!unzip -q fruits.zip -d content

!ls

!pip install --upgrade keras
!pip install tensorflow

from keras.models import Sequential
from keras.layers import Conv2D
from keras.layers import MaxPooling2D
from keras.layers import Flatten
from keras.layers import Dense
from keras.layers import Dropout
from keras.preprocessing import image
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image



no_of_classes = 131

# Initialising the CNN
classifier = Sequential()

# Step 1 - Convolution
classifier.add(Conv2D(32, (5, 5), input_shape = (100, 100, 3), activation = 'relu'))

# Step 2 - Pooling
classifier.add(MaxPooling2D(pool_size = (2, 2), strides= 2))

# Adding a convolutional layer
classifier.add(Conv2D(64, (5, 5), activation = 'relu'))
classifier.add(MaxPooling2D(pool_size = (2, 2), strides= 2))

# Adding a second convolutional layer
classifier.add(Conv2D(128, (5, 5), activation = 'relu'))
classifier.add(MaxPooling2D(pool_size = (2, 2), strides= 2))

# Adding a third convolutional layer
classifier.add(Conv2D(256, (5, 5), activation = 'relu'))
classifier.add(Dropout(0.25))
classifier.add(MaxPooling2D(pool_size = (2, 2), strides= 2))

# Step 3 - Flattening
classifier.add(Flatten())

# Step 4 - Full connection
classifier.add(Dense(units = 128, activation = 'relu'))
classifier.add(Dropout(0.25))
classifier.add(Dense(units = 128, activation = 'relu'))
classifier.add(Dropout(0.25))
classifier.add(Dense(units = no_of_classes, activation = 'softmax'))

# Compiling the CNN
classifier.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])

# Part 2 - Fitting the CNN to the images

from keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(rescale = 1./255,
                                   shear_range = 0.2,
                                   zoom_range = 0.2,
                                   horizontal_flip = True)

test_datagen = ImageDataGenerator(rescale = 1./255)

Training = train_datagen.flow_from_directory('/content/fruits-360/Training',
                                                 target_size = (100, 100),
                                                 batch_size = 32,
                                                 class_mode = 'categorical')

Validation = test_datagen.flow_from_directory('/content/fruits-360/Test',
                                            target_size = (100, 100),
                                            batch_size = 32,
                                            class_mode = 'categorical')

import os
os.walk('/content/fruits-360')
[x[0] for x in os.walk('/content/fruits-360/Training')]

"""# New Section"""

history = classifier.fit_generator(Training,
                         
                         epochs = 25,
                         verbose = 1,
                         validation_data = Validation,
                         
                         )
classifier.summary()

history.history['accuracy']

#Plotting graphs
acc = history.history['accuracy']
val_acc = history.history['val_accuracy']
loss = history.history['loss']
val_loss = history.history['val_loss']
 
epochs = range(len(acc))
 
plt.plot(epochs, acc, 'b', label='Training accuracy')
plt.plot(epochs, val_acc, 'r', label='Validation accuracy')
plt.title('Training and validation accuracy')
plt.legend()
 
plt.figure()
 
plt.plot(epochs, loss, 'b', label='Training loss')
plt.plot(epochs, val_loss, 'r', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
 
plt.show()

name_of_classes = [ ' Pitahaya Red', ' Apple Golden 1',' Beetroot', ' Pear Abate', ' Mulberry', ' Apple Golden 2', ' Pineapple', ' Pear Monster', ' Pear Red', ' Orange', ' Cherry Wax Red', ' Pear', ' Pear Stone', ' Banana', ' Limes', ' Apple Red Yellow 2', ' Apple Crimson Snow', ' Plum', ' Cactus fruit', ' Grape White 2', ' Pear 2', ' Pineapple Mini', ' Cantaloupe 2', ' Carambula', ' Cherry 1', ' Tamarillo', ' Papaya', ' Lemon',' Cherry 2', ' Tomato 2', ' Mandarine', ' Pomelo Sweetie', ' Pepper Red', ' Hazelnut', ' Nut Pecan', ' Physalis', ' Grape White 3', ' Grape White',' Cherry Wax Yellow', ' Apple Granny Smith', ' Tangelo', ' Kohlrabi', ' Melon Piel de Sapo', ' Cauliflower', ' Strawberry Wedge', ' Potato Sweet',' Apple Red 2', ' Apple Red 3', ' Grapefruit White', ' Avocado ripe', ' Cantaloupe 1', ' Tomato Maroon', ' Tomato Heart', ' Grape Blue', ' Walnut',' Grape Pink', ' Maracuja', ' Tomato Yellow', ' Lychee', ' Apple Red 1', ' Nectarine Flat', ' Pear Forelle', ' Tomato 4', ' Cherry Rainier',' Grapefruit Pink',' Peach',' Dates',' Mango Red',' Raspberry', ' Fig', ' Strawberry', ' Apple Pink Lady', ' Kumquats', ' Plum 3', ' Plum 2', ' Lemon Meyer', ' Pepper Yellow', ' Kaki', ' Tomato 1', ' Pepper Orange', ' Apple Golden 3', ' Potato White', ' Physalis with Husk', ' Mango',' Huckleberry',' Cocos',' Redcurrant',' Tomato not Ripened',' Passion Fruit', ' Apple Braeburn', ' Potato Red Washed',' Salak',' Granadilla',' Tomato 3',' Peach Flat', ' Apple Red Delicious', ' Banana Lady Finger', ' Watermelon', ' Clementine', ' Corn', ' Onion Red Peeled',' Onion White', ' Rambutan', ' Guava', ' Potato Red', ' Cucumber Ripe 2', ' Pepper Green', ' Grape White 4', ' Blueberry', ' Tomato Cherry Red',' Pomegranate',' Cucumber Ripe',' Mangostan', ' Pear Kaiser', ' Onion Red', ' Ginger Root', ' Pear Williams', ' Avocado',' Pepino', ' Corn Husk',' Banana Red', ' Apple Red Yellow 1', ' Chestnut', ' Eggplant', ' Quince', ' Cherry Wax Black', ' Nectarine', ' Kiwi', ' Nut Forest', ' Apricot', ' Peach 2']

Training.class_indices

test_image = image.load_img('/content/apple.jpg',target_size = (100, 100))
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image, axis = 0 )
result = classifier.predict(test_image)
Training.class_indices
img = '/content/apple.jpg'
Image.open('/content/apple.jpg')

for i in range(no_of_classes):
    if (result[0][i] == 1.0):
        img_path = '/content/apple.jpg'
        print ('Predicted:',name_of_classes[i])
Image.open(img_path)

test_image = image.load_img('/content/apple red yellow.jpg', target_size = (100, 100))
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image, axis = 0)
result = classifier.predict(test_image)
Training.class_indices
img = '/content/apple red yellow.jpg'
Image.open(img)

for i in range(no_of_classes):
    if (result[0][i] == 1.0):
        img_path = '/content/apple red yellow.jpg'
        print ('Predicted:',name_of_classes[i])
Image.open(img_path)

test_image = image.load_img('/content/chestnut.jpg', target_size = (100, 100))
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image, axis = 0)
result = classifier.predict(test_image)
Training.class_indices
img = '/content/chestnut.jpg'
Image.open(img)


for i in range(no_of_classes):
    if (result[0][i] == 1.0):
        img_path = '/content/chestnut.jpg'
        print ('Predicted:',name_of_classes[i])
Image.open(img_path)

test_image = image.load_img('/content/banana.jpg', target_size = (100, 100))
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image, axis = 0)
result = classifier.predict(test_image)
Training.class_indices
img = '/content/banana.jpg'
Image.open(img)


for i in range(no_of_classes):
    if (result[0][0] == 1.0):
        img_path = '/content/banana.jpg'
        print ('Predicted:',name_of_classes[i])
Image.open(img_path)

test_image = image.load_img('/content/grapes.jpg', target_size = (100, 100))
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image, axis = 0)
result = classifier.predict(test_image)
Training.class_indices
img = '/content/grapes.jpg'
Image.open(img)


for i in range(no_of_classes):
    if (result[0][i] == 1.0):
        img_path = '/content/grapes.jpg'
        print ('Predicted:',name_of_classes[i])
Image.open(img_path)

test_image = image.load_img('/content/mango.jpg', target_size = (100, 100))
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image, axis = 0)
result = classifier.predict(test_image)
Training.class_indices
img = '/content/mango.jpg'
Image.open(img)


for i in range(no_of_classes):
    if (result[0][i] == 1.0):
        img_path = '/content/mango.jpg'
        print ('Predicted:',name_of_classes[i])
Image.open(img_path)