форкнуто от main/is_dnn
Вы не можете выбрать более 25 тем
Темы должны начинаться с буквы или цифры, могут содержать дефисы(-) и должны содержать не более 35 символов.
436 KiB
436 KiB
# импорт модулей
import os
os.chdir('/content/drive/MyDrive/Colab Notebooks/is_lab3')
from tensorflow import keras
from tensorflow.keras import layers
from tensorflow.keras.models import Sequential
import matplotlib.pyplot as plt
import numpy as np
from sklearn.metrics import classification_report, confusion_matrix
from sklearn.metrics import ConfusionMatrixDisplay# загрузка датасета
from keras.datasets import mnist
(X_train, y_train), (X_test, y_test) = mnist.load_data()Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz
11490434/11490434 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
# создание своего разбиения датасета
from sklearn.model_selection import train_test_split
# объединяем в один набор
X = np.concatenate((X_train, X_test))
y = np.concatenate((y_train, y_test))
# разбиваем по вариантам
X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size = 10000,
train_size = 60000,
random_state = 11)
# вывод размерностей
print('Shape of X train:', X_train.shape)
print('Shape of y train:', y_train.shape)
print('Shape of X test:', X_test.shape)
print('Shape of y test:', y_test.shape)Shape of X train: (60000, 28, 28)
Shape of y train: (60000,)
Shape of X test: (10000, 28, 28)
Shape of y test: (10000,)
# Зададим параметры данных и модели
num_classes = 10
input_shape = (28, 28, 1)
# Приведение входных данных к диапазону [0, 1]
X_train = X_train / 255
X_test = X_test / 255
# Расширяем размерность входных данных, чтобы каждое изображение имело
# размерность (высота, ширина, количество каналов)
X_train = np.expand_dims(X_train, -1)
X_test = np.expand_dims(X_test, -1)
print('Shape of transformed X train:', X_train.shape)
print('Shape of transformed X test:', X_test.shape)
# переведем метки в one-hot
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)
print('Shape of transformed y train:', y_train.shape)
print('Shape of transformed y test:', y_test.shape)Shape of transformed X train: (60000, 28, 28, 1)
Shape of transformed X test: (10000, 28, 28, 1)
Shape of transformed y train: (60000, 10)
Shape of transformed y test: (10000, 10)
# создаем модель
model = Sequential()
model.add(layers.Conv2D(32, kernel_size=(3, 3), activation="relu", input_shape=input_shape))
model.add(layers.MaxPooling2D(pool_size=(2, 2)))
model.add(layers.Conv2D(64, kernel_size=(3, 3), activation="relu"))
model.add(layers.MaxPooling2D(pool_size=(2, 2)))
model.add(layers.Dropout(0.5))
model.add(layers.Flatten())
model.add(layers.Dense(num_classes, activation="softmax"))
model.summary()/usr/local/lib/python3.12/dist-packages/keras/src/layers/convolutional/base_conv.py:113: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
super().__init__(activity_regularizer=activity_regularizer, **kwargs)
Model: "sequential"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩ │ conv2d (Conv2D) │ (None, 26, 26, 32) │ 320 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d (MaxPooling2D) │ (None, 13, 13, 32) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_1 (Conv2D) │ (None, 11, 11, 64) │ 18,496 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d_1 (MaxPooling2D) │ (None, 5, 5, 64) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout (Dropout) │ (None, 5, 5, 64) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ flatten (Flatten) │ (None, 1600) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense (Dense) │ (None, 10) │ 16,010 │ └─────────────────────────────────┴────────────────────────┴───────────────┘
Total params: 34,826 (136.04 KB)
Trainable params: 34,826 (136.04 KB)
Non-trainable params: 0 (0.00 B)
# компилируем и обучаем модель
batch_size = 512
epochs = 15
model.compile(loss="categorical_crossentropy", optimizer="adam", metrics=["accuracy"])
model.fit(X_train, y_train, batch_size=batch_size, epochs=epochs, validation_split=0.1)Epoch 1/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 10s 49ms/step - accuracy: 0.5927 - loss: 1.3038 - val_accuracy: 0.9465 - val_loss: 0.1854
Epoch 2/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 11ms/step - accuracy: 0.9386 - loss: 0.2061 - val_accuracy: 0.9667 - val_loss: 0.1135
Epoch 3/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9579 - loss: 0.1391 - val_accuracy: 0.9725 - val_loss: 0.0891
Epoch 4/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9674 - loss: 0.1107 - val_accuracy: 0.9753 - val_loss: 0.0769
Epoch 5/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9722 - loss: 0.0937 - val_accuracy: 0.9780 - val_loss: 0.0684
Epoch 6/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9751 - loss: 0.0821 - val_accuracy: 0.9798 - val_loss: 0.0631
Epoch 7/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9770 - loss: 0.0751 - val_accuracy: 0.9805 - val_loss: 0.0591
Epoch 8/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9801 - loss: 0.0662 - val_accuracy: 0.9803 - val_loss: 0.0564
Epoch 9/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9808 - loss: 0.0616 - val_accuracy: 0.9810 - val_loss: 0.0543
Epoch 10/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 12ms/step - accuracy: 0.9816 - loss: 0.0602 - val_accuracy: 0.9837 - val_loss: 0.0494
Epoch 11/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 11ms/step - accuracy: 0.9835 - loss: 0.0530 - val_accuracy: 0.9838 - val_loss: 0.0481
Epoch 12/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 11ms/step - accuracy: 0.9845 - loss: 0.0503 - val_accuracy: 0.9837 - val_loss: 0.0450
Epoch 13/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9842 - loss: 0.0494 - val_accuracy: 0.9847 - val_loss: 0.0446
Epoch 14/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9857 - loss: 0.0471 - val_accuracy: 0.9843 - val_loss: 0.0452
Epoch 15/15
106/106 ━━━━━━━━━━━━━━━━━━━━ 1s 10ms/step - accuracy: 0.9856 - loss: 0.0476 - val_accuracy: 0.9840 - val_loss: 0.0450
<keras.src.callbacks.history.History at 0x7ce8ac8198e0># Оценка качества работы модели на тестовых данных
scores = model.evaluate(X_test, y_test)
print('Loss on test data:', scores[0])
print('Accuracy on test data:', scores[1])313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.9860 - loss: 0.0413
Loss on test data: 0.041157860308885574
Accuracy on test data: 0.9873999953269958
# вывод двух тестовых изображений и результатов распознавания
for n in [3,26]:
result = model.predict(X_test[n:n+1])
print('NN output:', result)
plt.imshow(X_test[n].reshape(28,28), cmap=plt.get_cmap('gray'))
plt.show()
print('Real mark: ', np.argmax(y_test[n]))
print('NN answer: ', np.argmax(result))1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 743ms/step
NN output: [[2.4183887e-08 3.2334963e-08 4.4088711e-08 4.4678579e-08 6.3559483e-08
2.0533466e-06 9.9999738e-01 1.5979442e-13 3.4749868e-07 7.8227762e-13]]
Real mark: 6
NN answer: 6
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step
NN output: [[4.9824855e-08 9.9986720e-01 5.8547844e-06 4.6495694e-07 7.3325129e-05
4.3212961e-07 4.2898904e-08 1.7418186e-05 3.4761622e-05 5.2838482e-07]]
Real mark: 1
NN answer: 1
# истинные метки классов
true_labels = np.argmax(y_test, axis=1)
# предсказанные метки классов
predicted_labels = np.argmax(model.predict(X_test), axis=1)
# отчет о качестве классификации
print(classification_report(true_labels, predicted_labels))
# вычисление матрицы ошибок
conf_matrix = confusion_matrix(true_labels, predicted_labels)
# отрисовка матрицы ошибок в виде "тепловой карты"
display = ConfusionMatrixDisplay(confusion_matrix=conf_matrix)
display.plot()
plt.show()313/313 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step
precision recall f1-score support
0 0.99 1.00 0.99 1018
1 0.99 0.99 0.99 1098
2 0.99 0.98 0.98 1010
3 0.99 0.98 0.98 992
4 0.99 0.98 0.99 998
5 0.99 0.99 0.99 951
6 0.99 0.99 0.99 933
7 0.98 0.99 0.98 1010
8 0.99 0.99 0.99 960
9 0.98 0.99 0.98 1030
accuracy 0.99 10000
macro avg 0.99 0.99 0.99 10000
weighted avg 0.99 0.99 0.99 10000
# загрузка собственного изображения
from PIL import Image
for name_image in ['2.png', '5.png']:
file_data = Image.open(name_image)
file_data = file_data.convert('L') # перевод в градации серого
test_img = np.array(file_data)
# вывод собственного изображения
plt.imshow(test_img, cmap=plt.get_cmap('gray'))
plt.show()
# предобработка
test_img = test_img / 255
test_img = np.reshape(test_img, (1,28,28,1))
# распознавание
result = model.predict(test_img)
print('I think it\'s', np.argmax(result))
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 44ms/step
I think it's 2
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 41ms/step
I think it's 5
model_lr1 = keras.models.load_model("best_model.keras")
model_lr1.summary()Model: "sequential_7"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩ │ dense_14 (Dense) │ (None, 100) │ 78,500 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense_15 (Dense) │ (None, 100) │ 10,100 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense_16 (Dense) │ (None, 10) │ 1,010 │ └─────────────────────────────────┴────────────────────────┴───────────────┘
Total params: 89,612 (350.05 KB)
Trainable params: 89,610 (350.04 KB)
Non-trainable params: 0 (0.00 B)
Optimizer params: 2 (12.00 B)
# развернем каждое изображение 28*28 в вектор 784
X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size = 10000,
train_size = 60000,
random_state = 11)
num_pixels = X_train.shape[1] * X_train.shape[2]
X_train = X_train.reshape(X_train.shape[0], num_pixels) / 255
X_test = X_test.reshape(X_test.shape[0], num_pixels) / 255
print('Shape of transformed X train:', X_train.shape)
print('Shape of transformed X train:', X_test.shape)
# переведем метки в one-hot
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)
print('Shape of transformed y train:', y_train.shape)
print('Shape of transformed y test:', y_test.shape)Shape of transformed X train: (60000, 784)
Shape of transformed X train: (10000, 784)
Shape of transformed y train: (60000, 10)
Shape of transformed y test: (10000, 10)
# Оценка качества работы модели на тестовых данных
scores = model_lr1.evaluate(X_test, y_test)
print('Loss on test data:', scores[0])
print('Accuracy on test data:', scores[1])313/313 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.9406 - loss: 0.2055
Loss on test data: 0.1955890655517578
Accuracy on test data: 0.9426000118255615
# загрузка датасета
from keras.datasets import cifar10
(X_train, y_train), (X_test, y_test) = cifar10.load_data()Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
170498071/170498071 ━━━━━━━━━━━━━━━━━━━━ 4s 0us/step
# создание своего разбиения датасета
# объединяем в один набор
X = np.concatenate((X_train, X_test))
y = np.concatenate((y_train, y_test))
# разбиваем по вариантам
X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size = 10000,
train_size = 50000,
random_state = 11)
# вывод размерностей
print('Shape of X train:', X_train.shape)
print('Shape of y train:', y_train.shape)
print('Shape of X test:', X_test.shape)
print('Shape of y test:', y_test.shape)Shape of X train: (50000, 32, 32, 3)
Shape of y train: (50000, 1)
Shape of X test: (10000, 32, 32, 3)
Shape of y test: (10000, 1)
class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer',
'dog', 'frog', 'horse', 'ship', 'truck']
plt.figure(figsize=(10,10))
for i in range(25):
plt.subplot(5,5,i+1)
plt.xticks([])
plt.yticks([])
plt.grid(False)
plt.imshow(X_train[i])
plt.xlabel(class_names[y_train[i][0]])
plt.show()
# Зададим параметры данных и модели
num_classes = 10
input_shape = (32, 32, 3)
# Приведение входных данных к диапазону [0, 1]
X_train = X_train / 255
X_test = X_test / 255
print('Shape of transformed X train:', X_train.shape)
print('Shape of transformed X test:', X_test.shape)
# переведем метки в one-hot
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)
print('Shape of transformed y train:', y_train.shape)
print('Shape of transformed y test:', y_test.shape)Shape of transformed X train: (50000, 32, 32, 3)
Shape of transformed X test: (10000, 32, 32, 3)
Shape of transformed y train: (50000, 10)
Shape of transformed y test: (10000, 10)
# создаем модель
model = Sequential()
# Блок 1
model.add(layers.Conv2D(32, (3, 3), padding="same",
activation="relu", input_shape=input_shape))
model.add(layers.BatchNormalization())
model.add(layers.Conv2D(32, (3, 3), padding="same", activation="relu"))
model.add(layers.BatchNormalization())
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Dropout(0.25))
# Блок 2
model.add(layers.Conv2D(64, (3, 3), padding="same", activation="relu"))
model.add(layers.BatchNormalization())
model.add(layers.Conv2D(64, (3, 3), padding="same", activation="relu"))
model.add(layers.BatchNormalization())
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Dropout(0.25))
# Блок 3
model.add(layers.Conv2D(128, (3, 3), padding="same", activation="relu"))
model.add(layers.BatchNormalization())
model.add(layers.Conv2D(128, (3, 3), padding="same", activation="relu"))
model.add(layers.BatchNormalization())
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Dropout(0.4))
model.add(layers.Flatten())
model.add(layers.Dense(128, activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(num_classes, activation="softmax"))
model.summary()/usr/local/lib/python3.12/dist-packages/keras/src/layers/convolutional/base_conv.py:113: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.
super().__init__(activity_regularizer=activity_regularizer, **kwargs)
Model: "sequential_1"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩ │ conv2d_2 (Conv2D) │ (None, 32, 32, 32) │ 896 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization │ (None, 32, 32, 32) │ 128 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_3 (Conv2D) │ (None, 32, 32, 32) │ 9,248 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_1 │ (None, 32, 32, 32) │ 128 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d_2 (MaxPooling2D) │ (None, 16, 16, 32) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_1 (Dropout) │ (None, 16, 16, 32) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_4 (Conv2D) │ (None, 16, 16, 64) │ 18,496 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_2 │ (None, 16, 16, 64) │ 256 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_5 (Conv2D) │ (None, 16, 16, 64) │ 36,928 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_3 │ (None, 16, 16, 64) │ 256 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d_3 (MaxPooling2D) │ (None, 8, 8, 64) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_2 (Dropout) │ (None, 8, 8, 64) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_6 (Conv2D) │ (None, 8, 8, 128) │ 73,856 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_4 │ (None, 8, 8, 128) │ 512 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_7 (Conv2D) │ (None, 8, 8, 128) │ 147,584 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_5 │ (None, 8, 8, 128) │ 512 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d_4 (MaxPooling2D) │ (None, 4, 4, 128) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_3 (Dropout) │ (None, 4, 4, 128) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ flatten_1 (Flatten) │ (None, 2048) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense_1 (Dense) │ (None, 128) │ 262,272 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_4 (Dropout) │ (None, 128) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense_2 (Dense) │ (None, 10) │ 1,290 │ └─────────────────────────────────┴────────────────────────┴───────────────┘
Total params: 552,362 (2.11 MB)
Trainable params: 551,466 (2.10 MB)
Non-trainable params: 896 (3.50 KB)
# компилируем и обучаем модель
batch_size = 64
epochs = 50
model.compile(loss="categorical_crossentropy", optimizer="adam", metrics=["accuracy"])
model.fit(X_train, y_train, batch_size=batch_size, epochs=epochs, validation_split=0.1)Epoch 1/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 28s 22ms/step - accuracy: 0.2674 - loss: 2.1039 - val_accuracy: 0.2834 - val_loss: 2.3722
Epoch 2/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.4822 - loss: 1.4320 - val_accuracy: 0.5832 - val_loss: 1.1490
Epoch 3/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.5803 - loss: 1.1857 - val_accuracy: 0.6454 - val_loss: 1.0278
Epoch 4/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.6493 - loss: 1.0135 - val_accuracy: 0.6832 - val_loss: 0.9214
Epoch 5/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.6807 - loss: 0.9211 - val_accuracy: 0.6854 - val_loss: 0.9277
Epoch 6/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.7099 - loss: 0.8405 - val_accuracy: 0.7246 - val_loss: 0.8032
Epoch 7/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.7338 - loss: 0.7774 - val_accuracy: 0.7444 - val_loss: 0.7713
Epoch 8/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.7585 - loss: 0.7138 - val_accuracy: 0.7430 - val_loss: 0.7700
Epoch 9/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.7723 - loss: 0.6724 - val_accuracy: 0.7964 - val_loss: 0.6089
Epoch 10/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.7834 - loss: 0.6431 - val_accuracy: 0.7510 - val_loss: 0.7581
Epoch 11/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 10ms/step - accuracy: 0.7938 - loss: 0.6043 - val_accuracy: 0.7412 - val_loss: 0.8413
Epoch 12/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8078 - loss: 0.5686 - val_accuracy: 0.8084 - val_loss: 0.5824
Epoch 13/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8135 - loss: 0.5516 - val_accuracy: 0.7962 - val_loss: 0.6150
Epoch 14/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 9s 9ms/step - accuracy: 0.8224 - loss: 0.5198 - val_accuracy: 0.8146 - val_loss: 0.5601
Epoch 15/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8266 - loss: 0.5162 - val_accuracy: 0.7736 - val_loss: 0.7025
Epoch 16/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.8375 - loss: 0.4846 - val_accuracy: 0.8144 - val_loss: 0.5579
Epoch 17/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8414 - loss: 0.4571 - val_accuracy: 0.8214 - val_loss: 0.5381
Epoch 18/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.8463 - loss: 0.4582 - val_accuracy: 0.8282 - val_loss: 0.5478
Epoch 19/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.8508 - loss: 0.4418 - val_accuracy: 0.8260 - val_loss: 0.5413
Epoch 20/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.8584 - loss: 0.4141 - val_accuracy: 0.8200 - val_loss: 0.5675
Epoch 21/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.8642 - loss: 0.4024 - val_accuracy: 0.8288 - val_loss: 0.5301
Epoch 22/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8635 - loss: 0.4004 - val_accuracy: 0.8352 - val_loss: 0.5078
Epoch 23/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.8689 - loss: 0.3831 - val_accuracy: 0.8342 - val_loss: 0.5113
Epoch 24/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8697 - loss: 0.3863 - val_accuracy: 0.8300 - val_loss: 0.5583
Epoch 25/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.8737 - loss: 0.3708 - val_accuracy: 0.8222 - val_loss: 0.5647
Epoch 26/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8779 - loss: 0.3573 - val_accuracy: 0.8322 - val_loss: 0.5362
Epoch 27/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.8834 - loss: 0.3445 - val_accuracy: 0.8436 - val_loss: 0.5163
Epoch 28/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8823 - loss: 0.3487 - val_accuracy: 0.8352 - val_loss: 0.5274
Epoch 29/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.8852 - loss: 0.3349 - val_accuracy: 0.8370 - val_loss: 0.5532
Epoch 30/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8873 - loss: 0.3230 - val_accuracy: 0.8154 - val_loss: 0.6128
Epoch 31/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.8886 - loss: 0.3198 - val_accuracy: 0.8448 - val_loss: 0.4880
Epoch 32/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8949 - loss: 0.3061 - val_accuracy: 0.8476 - val_loss: 0.5057
Epoch 33/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 6s 9ms/step - accuracy: 0.8946 - loss: 0.3115 - val_accuracy: 0.8492 - val_loss: 0.5195
Epoch 34/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8976 - loss: 0.3023 - val_accuracy: 0.8424 - val_loss: 0.5272
Epoch 35/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8988 - loss: 0.2896 - val_accuracy: 0.8524 - val_loss: 0.5089
Epoch 36/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9006 - loss: 0.2894 - val_accuracy: 0.8420 - val_loss: 0.5439
Epoch 37/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.9030 - loss: 0.2828 - val_accuracy: 0.8386 - val_loss: 0.5211
Epoch 38/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9062 - loss: 0.2717 - val_accuracy: 0.8482 - val_loss: 0.5060
Epoch 39/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.9056 - loss: 0.2753 - val_accuracy: 0.8426 - val_loss: 0.5156
Epoch 40/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.9081 - loss: 0.2687 - val_accuracy: 0.8476 - val_loss: 0.4856
Epoch 41/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9048 - loss: 0.2731 - val_accuracy: 0.8490 - val_loss: 0.5141
Epoch 42/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.9089 - loss: 0.2662 - val_accuracy: 0.8430 - val_loss: 0.5272
Epoch 43/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9135 - loss: 0.2573 - val_accuracy: 0.8586 - val_loss: 0.4969
Epoch 44/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.9108 - loss: 0.2583 - val_accuracy: 0.8596 - val_loss: 0.4902
Epoch 45/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9168 - loss: 0.2432 - val_accuracy: 0.8526 - val_loss: 0.4931
Epoch 46/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 10s 9ms/step - accuracy: 0.9146 - loss: 0.2482 - val_accuracy: 0.8434 - val_loss: 0.5511
Epoch 47/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9135 - loss: 0.2529 - val_accuracy: 0.8508 - val_loss: 0.5254
Epoch 48/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 9ms/step - accuracy: 0.9200 - loss: 0.2337 - val_accuracy: 0.8572 - val_loss: 0.4973
Epoch 49/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9160 - loss: 0.2438 - val_accuracy: 0.8424 - val_loss: 0.5488
Epoch 50/50
704/704 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.9160 - loss: 0.2469 - val_accuracy: 0.8572 - val_loss: 0.5109
<keras.src.callbacks.history.History at 0x7ce87461dd00># Оценка качества работы модели на тестовых данных
scores = model.evaluate(X_test, y_test)
print('Loss on test data:', scores[0])
print('Accuracy on test data:', scores[1])313/313 ━━━━━━━━━━━━━━━━━━━━ 7s 10ms/step - accuracy: 0.8615 - loss: 0.4696
Loss on test data: 0.4909549355506897
Accuracy on test data: 0.857200026512146
# вывод двух тестовых изображений и результатов распознавания
for n in [3,15]:
result = model.predict(X_test[n:n+1])
print('NN output:', result)
plt.imshow(X_test[n].reshape(32,32,3), cmap=plt.get_cmap('gray'))
plt.show()
print('Real mark: ', np.argmax(y_test[n]))
print('NN answer: ', np.argmax(result))1/1 ━━━━━━━━━━━━━━━━━━━━ 2s 2s/step
NN output: [[3.6980788e-04 6.5029941e-05 4.4359174e-02 8.4741175e-02 1.3959331e-02
8.4197390e-01 5.7383772e-04 1.3647835e-02 1.1692162e-05 2.9820076e-04]]
Real mark: 4
NN answer: 5
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 101ms/step
NN output: [[1.0492409e-07 3.9376263e-12 9.9999964e-01 1.5196024e-08 1.3079713e-07
4.7579718e-08 3.4805732e-09 1.2734243e-07 6.2821698e-10 1.3412522e-11]]
Real mark: 2
NN answer: 2
# истинные метки классов
true_labels = np.argmax(y_test, axis=1)
# предсказанные метки классов
predicted_labels = np.argmax(model.predict(X_test), axis=1)
# отчет о качестве классификации
print(classification_report(true_labels, predicted_labels, target_names=class_names))
# вычисление матрицы ошибок
conf_matrix = confusion_matrix(true_labels, predicted_labels)
# отрисовка матрицы ошибок в виде "тепловой карты"
fig, ax = plt.subplots(figsize=(6, 6))
disp = ConfusionMatrixDisplay(confusion_matrix=conf_matrix,display_labels=class_names)
disp.plot(ax=ax, xticks_rotation=45) # поворот подписей по X и приятная палитра
plt.tight_layout() # чтобы всё влезло
plt.show()313/313 ━━━━━━━━━━━━━━━━━━━━ 4s 10ms/step
precision recall f1-score support
airplane 0.89 0.86 0.87 1000
automobile 0.94 0.93 0.93 1019
bird 0.80 0.84 0.82 972
cat 0.78 0.66 0.72 1014
deer 0.84 0.84 0.84 980
dog 0.76 0.81 0.79 1051
frog 0.90 0.87 0.89 1043
horse 0.89 0.89 0.89 1018
ship 0.91 0.94 0.92 945
truck 0.86 0.94 0.90 958
accuracy 0.86 10000
macro avg 0.86 0.86 0.86 10000
weighted avg 0.86 0.86 0.86 10000
