
import numpy as np
import tensorflow as tf
from tensorflow.keras.datasets import cifar10
import matplotlib.pyplot as plt

from tensorflow.keras.utils import to_categorical
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Flatten, Dense
from tensorflow.keras.optimizers import SGD, Adam
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Dropout, BatchNormalization, Activation, GlobalAveragePooling2D
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau


(train_images, train_labels), (test_images, test_labels) = cifar10.load_data()


print(train_images.shape, train_labels.shape)
print(test_images.shape, test_labels.shape)
print(train_images[0].shape)

(50000, 32, 32, 3) (50000, 1)
(10000, 32, 32, 3) (10000, 1)
(32, 32, 3)


class_names = ["airplane", "automobile", "bird", "cat", "deer", "dog", "frog", "horse", "ship", "truck"]


train_labels[:10]

array([[6],
       [9],
       [9],
       [4],
       [1],
       [1],
       [2],
       [7],
       [8],
       [3]], dtype=uint8)


train_labels = np.int32(train_labels.flatten())


figure, axes = plt.subplots(figsize=(22, 10), nrows=3, ncols=8)

for i in range(3):
    for j in range(8):
        axes[i][j].imshow(train_images[(i*8)+j])
        axes[i][j].set_title(class_names[train_labels[(i*8)+j]])


train_images = np.array(train_images/255.0, dtype=np.float32)
train_labels = np.array(train_labels, dtype=np.float32)
test_images = np.array(test_images/255.0, dtype=np.float32)
test_labels = np.array(test_labels, dtype=np.float32)

train_oh_labels = to_categorical(train_labels)
test_oh_labels = to_categorical(test_labels)


test_oh_labels

array([[0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 1., 0.],
       [0., 0., 0., ..., 0., 1., 0.],
       ...,
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 1., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 1., 0., 0.]], dtype=float32)


def build_model():
    tf.keras.backend.clear_session()

    input_tensor = Input(shape=(32, 32, 3))
    x = Conv2D(filters=32, kernel_size=(3, 3), strides=1, padding="same")(input_tensor)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=32, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)

    x = Conv2D(filters=64, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=64, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)
                    
    x = Conv2D(filters=128, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=128, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)                 
                    
    x = Flatten()(x)
    x = Dropout(rate=0.5)(x)                 
    x = Dense(300, activation="relu")(x)
    x = Dropout(rate=0.3)(x)
    x = Dense(100, activation="relu")(x)
    x = Dropout(rate=0.3)(x)
    output = Dense(10, activation="softmax")(x)

    model = Model(inputs=input_tensor, outputs=output)
    return model


model = build_model()


model.compile(optimizer=Adam(0.001), loss="categorical_crossentropy", metrics=["accuracy"])


batch_size = 32
history = model.fit(x=train_images, y=train_oh_labels, batch_size=batch_size, epochs=30, validation_split=0.2)

Epoch 1/30
1250/1250 [==============================] - 57s 24ms/step - loss: 1.7036 - accuracy: 0.3796 - val_loss: 1.4955 - val_accuracy: 0.4691
Epoch 2/30
1250/1250 [==============================] - 26s 21ms/step - loss: 1.2347 - accuracy: 0.5646 - val_loss: 1.0512 - val_accuracy: 0.6251
Epoch 3/30
1250/1250 [==============================] - 26s 21ms/step - loss: 1.0222 - accuracy: 0.6457 - val_loss: 0.9102 - val_accuracy: 0.6883
Epoch 4/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.9004 - accuracy: 0.6948 - val_loss: 0.8284 - val_accuracy: 0.7161
Epoch 5/30
1250/1250 [==============================] - 26s 20ms/step - loss: 0.8060 - accuracy: 0.7269 - val_loss: 0.9674 - val_accuracy: 0.6810
Epoch 6/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.7394 - accuracy: 0.7510 - val_loss: 0.7857 - val_accuracy: 0.7357
Epoch 7/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.6772 - accuracy: 0.7714 - val_loss: 0.7322 - val_accuracy: 0.7436
Epoch 8/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.6254 - accuracy: 0.7914 - val_loss: 1.0344 - val_accuracy: 0.6743
Epoch 9/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.5869 - accuracy: 0.8044 - val_loss: 0.6846 - val_accuracy: 0.7713
Epoch 10/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.5391 - accuracy: 0.8213 - val_loss: 0.7362 - val_accuracy: 0.7537
Epoch 11/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.5051 - accuracy: 0.8321 - val_loss: 0.6491 - val_accuracy: 0.7922
Epoch 12/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.4700 - accuracy: 0.8437 - val_loss: 0.6870 - val_accuracy: 0.7793
Epoch 13/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.4364 - accuracy: 0.8536 - val_loss: 0.5211 - val_accuracy: 0.8246
Epoch 14/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.4097 - accuracy: 0.8644 - val_loss: 0.6611 - val_accuracy: 0.7942
Epoch 15/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3902 - accuracy: 0.8710 - val_loss: 0.6795 - val_accuracy: 0.7842
Epoch 16/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3661 - accuracy: 0.8784 - val_loss: 0.6497 - val_accuracy: 0.7996
Epoch 17/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3452 - accuracy: 0.8855 - val_loss: 0.5536 - val_accuracy: 0.8268
Epoch 18/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3194 - accuracy: 0.8937 - val_loss: 0.6322 - val_accuracy: 0.8030
Epoch 19/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3083 - accuracy: 0.8970 - val_loss: 0.5713 - val_accuracy: 0.8224
Epoch 20/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2920 - accuracy: 0.9038 - val_loss: 0.5717 - val_accuracy: 0.8419
Epoch 21/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2845 - accuracy: 0.9060 - val_loss: 0.5698 - val_accuracy: 0.8272
Epoch 22/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2658 - accuracy: 0.9104 - val_loss: 0.5669 - val_accuracy: 0.8341
Epoch 23/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2527 - accuracy: 0.9172 - val_loss: 0.6116 - val_accuracy: 0.8264
Epoch 24/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2458 - accuracy: 0.9202 - val_loss: 0.7002 - val_accuracy: 0.8146
Epoch 25/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2270 - accuracy: 0.9248 - val_loss: 0.6205 - val_accuracy: 0.8215
Epoch 26/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2257 - accuracy: 0.9244 - val_loss: 0.6134 - val_accuracy: 0.8262
Epoch 27/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2112 - accuracy: 0.9307 - val_loss: 0.6482 - val_accuracy: 0.8279
Epoch 28/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2083 - accuracy: 0.9312 - val_loss: 0.6227 - val_accuracy: 0.8309
Epoch 29/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2052 - accuracy: 0.9334 - val_loss: 0.6521 - val_accuracy: 0.8268
Epoch 30/30
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1893 - accuracy: 0.9380 - val_loss: 0.6303 - val_accuracy: 0.8271


plt.plot(history.history["accuracy"], label="train")
plt.plot(history.history["val_accuracy"], label="valid")
plt.legend()

<matplotlib.legend.Legend at 0x7f3110b386d0>


evaluate_accuracy = model.evaluate(test_images, test_oh_labels, batch_size=batch_size, verbose=1)

313/313 [==============================] - 3s 9ms/step - loss: 0.6434 - accuracy: 0.8238


model = build_model()
model.summary()

Model: "model"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 32, 32, 3)]       0         
_________________________________________________________________
conv2d (Conv2D)              (None, 32, 32, 32)        896       
_________________________________________________________________
batch_normalization (BatchNo (None, 32, 32, 32)        128       
_________________________________________________________________
activation (Activation)      (None, 32, 32, 32)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 32, 32, 32)        9248      
_________________________________________________________________
batch_normalization_1 (Batch (None, 32, 32, 32)        128       
_________________________________________________________________
activation_1 (Activation)    (None, 32, 32, 32)        0         
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 16, 16, 32)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 16, 16, 64)        18496     
_________________________________________________________________
batch_normalization_2 (Batch (None, 16, 16, 64)        256       
_________________________________________________________________
activation_2 (Activation)    (None, 16, 16, 64)        0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 16, 16, 64)        36928     
_________________________________________________________________
batch_normalization_3 (Batch (None, 16, 16, 64)        256       
_________________________________________________________________
activation_3 (Activation)    (None, 16, 16, 64)        0         
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 8, 8, 64)          0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 8, 8, 128)         73856     
_________________________________________________________________
batch_normalization_4 (Batch (None, 8, 8, 128)         512       
_________________________________________________________________
activation_4 (Activation)    (None, 8, 8, 128)         0         
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 8, 8, 128)         147584    
_________________________________________________________________
batch_normalization_5 (Batch (None, 8, 8, 128)         512       
_________________________________________________________________
activation_5 (Activation)    (None, 8, 8, 128)         0         
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 4, 4, 128)         0         
_________________________________________________________________
flatten (Flatten)            (None, 2048)              0         
_________________________________________________________________
dropout (Dropout)            (None, 2048)              0         
_________________________________________________________________
dense (Dense)                (None, 300)               614700    
_________________________________________________________________
dropout_1 (Dropout)          (None, 300)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 100)               30100     
_________________________________________________________________
dropout_2 (Dropout)          (None, 100)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 10)                1010      
=================================================================
Total params: 934,610
Trainable params: 933,714
Non-trainable params: 896
_________________________________________________________________


model.compile(optimizer=Adam(0.001), loss="categorical_crossentropy", metrics=["accuracy"])


checkpoint_cb = ModelCheckpoint("my_keras_model.h5", save_best_only=True, verbose=1)
early_stopping_cb = EarlyStopping(patience=10, restore_best_weights=True)
reducelr_cb = ReduceLROnPlateau(monitor="val_loss", factor=0.2, patience=5, mode="min", verbose=1)


batch_size = 32
history = model.fit(x=train_images, y=train_oh_labels, batch_size=batch_size, epochs=40, validation_split=0.2, 
                    callbacks=[checkpoint_cb, early_stopping_cb, reducelr_cb])

Epoch 1/40
1250/1250 [==============================] - 31s 24ms/step - loss: 1.7270 - accuracy: 0.3663 - val_loss: 1.6567 - val_accuracy: 0.4408

Epoch 00001: val_loss improved from inf to 1.65672, saving model to my_keras_model.h5
Epoch 2/40
1250/1250 [==============================] - 28s 23ms/step - loss: 1.2361 - accuracy: 0.5613 - val_loss: 1.1919 - val_accuracy: 0.5717

Epoch 00002: val_loss improved from 1.65672 to 1.19195, saving model to my_keras_model.h5
Epoch 3/40
1250/1250 [==============================] - 28s 22ms/step - loss: 1.0279 - accuracy: 0.6424 - val_loss: 1.0462 - val_accuracy: 0.6360

Epoch 00003: val_loss improved from 1.19195 to 1.04619, saving model to my_keras_model.h5
Epoch 4/40
1250/1250 [==============================] - 27s 22ms/step - loss: 0.9087 - accuracy: 0.6920 - val_loss: 0.8407 - val_accuracy: 0.7099

Epoch 00004: val_loss improved from 1.04619 to 0.84072, saving model to my_keras_model.h5
Epoch 5/40
1250/1250 [==============================] - 27s 22ms/step - loss: 0.8151 - accuracy: 0.7250 - val_loss: 0.7887 - val_accuracy: 0.7324

Epoch 00005: val_loss improved from 0.84072 to 0.78870, saving model to my_keras_model.h5
Epoch 6/40
1250/1250 [==============================] - 27s 22ms/step - loss: 0.7436 - accuracy: 0.7491 - val_loss: 0.9310 - val_accuracy: 0.6763

Epoch 00006: val_loss did not improve from 0.78870
Epoch 7/40
1250/1250 [==============================] - 27s 22ms/step - loss: 0.6743 - accuracy: 0.7742 - val_loss: 0.9520 - val_accuracy: 0.6706

Epoch 00007: val_loss did not improve from 0.78870
Epoch 8/40
1250/1250 [==============================] - 27s 22ms/step - loss: 0.6211 - accuracy: 0.7937 - val_loss: 0.7129 - val_accuracy: 0.7521

Epoch 00008: val_loss improved from 0.78870 to 0.71291, saving model to my_keras_model.h5
Epoch 9/40
1250/1250 [==============================] - 27s 21ms/step - loss: 0.5771 - accuracy: 0.8086 - val_loss: 0.6690 - val_accuracy: 0.7727

Epoch 00009: val_loss improved from 0.71291 to 0.66896, saving model to my_keras_model.h5
Epoch 10/40
1250/1250 [==============================] - 27s 22ms/step - loss: 0.5344 - accuracy: 0.8210 - val_loss: 0.6299 - val_accuracy: 0.7870

Epoch 00010: val_loss improved from 0.66896 to 0.62987, saving model to my_keras_model.h5
Epoch 11/40
1250/1250 [==============================] - 27s 22ms/step - loss: 0.4921 - accuracy: 0.8350 - val_loss: 0.6935 - val_accuracy: 0.7637

Epoch 00011: val_loss did not improve from 0.62987
Epoch 12/40
1250/1250 [==============================] - 27s 21ms/step - loss: 0.4654 - accuracy: 0.8456 - val_loss: 0.7862 - val_accuracy: 0.7489

Epoch 00012: val_loss did not improve from 0.62987
Epoch 13/40
1250/1250 [==============================] - 27s 21ms/step - loss: 0.4308 - accuracy: 0.8553 - val_loss: 0.5447 - val_accuracy: 0.8225

Epoch 00013: val_loss improved from 0.62987 to 0.54470, saving model to my_keras_model.h5
Epoch 14/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.4126 - accuracy: 0.8637 - val_loss: 0.5707 - val_accuracy: 0.8090

Epoch 00014: val_loss did not improve from 0.54470
Epoch 15/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3871 - accuracy: 0.8711 - val_loss: 0.5641 - val_accuracy: 0.8200

Epoch 00015: val_loss did not improve from 0.54470
Epoch 16/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3655 - accuracy: 0.8776 - val_loss: 0.5912 - val_accuracy: 0.8183

Epoch 00016: val_loss did not improve from 0.54470
Epoch 17/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3410 - accuracy: 0.8865 - val_loss: 0.6671 - val_accuracy: 0.7977

Epoch 00017: val_loss did not improve from 0.54470
Epoch 18/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.3278 - accuracy: 0.8899 - val_loss: 0.7426 - val_accuracy: 0.7871

Epoch 00018: val_loss did not improve from 0.54470

Epoch 00018: ReduceLROnPlateau reducing learning rate to 0.00020000000949949026.
Epoch 19/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.2311 - accuracy: 0.9234 - val_loss: 0.5059 - val_accuracy: 0.8529

Epoch 00019: val_loss improved from 0.54470 to 0.50591, saving model to my_keras_model.h5
Epoch 20/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1915 - accuracy: 0.9360 - val_loss: 0.4994 - val_accuracy: 0.8566

Epoch 00020: val_loss improved from 0.50591 to 0.49938, saving model to my_keras_model.h5
Epoch 21/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1746 - accuracy: 0.9405 - val_loss: 0.5337 - val_accuracy: 0.8538

Epoch 00021: val_loss did not improve from 0.49938
Epoch 22/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1589 - accuracy: 0.9470 - val_loss: 0.5283 - val_accuracy: 0.8558

Epoch 00022: val_loss did not improve from 0.49938
Epoch 23/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1458 - accuracy: 0.9506 - val_loss: 0.5554 - val_accuracy: 0.8500

Epoch 00023: val_loss did not improve from 0.49938
Epoch 24/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1421 - accuracy: 0.9526 - val_loss: 0.5563 - val_accuracy: 0.8537

Epoch 00024: val_loss did not improve from 0.49938
Epoch 25/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1328 - accuracy: 0.9551 - val_loss: 0.5500 - val_accuracy: 0.8589

Epoch 00025: val_loss did not improve from 0.49938

Epoch 00025: ReduceLROnPlateau reducing learning rate to 4.0000001899898055e-05.
Epoch 26/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1117 - accuracy: 0.9626 - val_loss: 0.5447 - val_accuracy: 0.8621

Epoch 00026: val_loss did not improve from 0.49938
Epoch 27/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1095 - accuracy: 0.9627 - val_loss: 0.5633 - val_accuracy: 0.8607

Epoch 00027: val_loss did not improve from 0.49938
Epoch 28/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1032 - accuracy: 0.9650 - val_loss: 0.5715 - val_accuracy: 0.8618

Epoch 00028: val_loss did not improve from 0.49938
Epoch 29/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.1016 - accuracy: 0.9650 - val_loss: 0.5802 - val_accuracy: 0.8604

Epoch 00029: val_loss did not improve from 0.49938
Epoch 30/40
1250/1250 [==============================] - 26s 21ms/step - loss: 0.0984 - accuracy: 0.9663 - val_loss: 0.5865 - val_accuracy: 0.8599

Epoch 00030: val_loss did not improve from 0.49938

Epoch 00030: ReduceLROnPlateau reducing learning rate to 8.000000525498762e-06.


plt.plot(history.history["accuracy"], label="train")
plt.plot(history.history["val_accuracy"], label="valid")
plt.legend()

<matplotlib.legend.Legend at 0x7f30c5e2d910>


model.evaluate(test_images, test_oh_labels, batch_size=batch_size, verbose=1)

313/313 [==============================] - 3s 9ms/step - loss: 0.5338 - accuracy: 0.8483

[0.5337657928466797, 0.8482999801635742]


def build_extended_model():
    tf.keras.backend.clear_session()

    input_tensor = Input(shape=(32, 32, 3))
    x = Conv2D(filters=32, kernel_size=(3, 3), strides=1, padding="same")(input_tensor)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=64, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)

    x = Conv2D(filters=128, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=128, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)
                    
    x = Conv2D(filters=256, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=256, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    # x = MaxPooling2D(pool_size=(2, 2))(x)                 

    x = Conv2D(filters=512, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)  
                    
    x = Flatten()(x)
    x = Dropout(rate=0.5)(x)                 
    x = Dense(300, activation="relu")(x)
    x = Dropout(rate=0.3)(x)
    x = Dense(100, activation="relu")(x)
    x = Dropout(rate=0.3)(x)
    output = Dense(10, activation="softmax")(x)

    model = Model(inputs=input_tensor, outputs=output)
    return model


model = build_extended_model()
model.summary()

Model: "model"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 32, 32, 3)]       0         
_________________________________________________________________
conv2d (Conv2D)              (None, 32, 32, 32)        896       
_________________________________________________________________
batch_normalization (BatchNo (None, 32, 32, 32)        128       
_________________________________________________________________
activation (Activation)      (None, 32, 32, 32)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 32, 32, 64)        18496     
_________________________________________________________________
batch_normalization_1 (Batch (None, 32, 32, 64)        256       
_________________________________________________________________
activation_1 (Activation)    (None, 32, 32, 64)        0         
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 16, 16, 64)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 16, 16, 128)       73856     
_________________________________________________________________
batch_normalization_2 (Batch (None, 16, 16, 128)       512       
_________________________________________________________________
activation_2 (Activation)    (None, 16, 16, 128)       0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 16, 16, 128)       147584    
_________________________________________________________________
batch_normalization_3 (Batch (None, 16, 16, 128)       512       
_________________________________________________________________
activation_3 (Activation)    (None, 16, 16, 128)       0         
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 8, 8, 128)         0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 8, 8, 256)         295168    
_________________________________________________________________
batch_normalization_4 (Batch (None, 8, 8, 256)         1024      
_________________________________________________________________
activation_4 (Activation)    (None, 8, 8, 256)         0         
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 8, 8, 256)         590080    
_________________________________________________________________
batch_normalization_5 (Batch (None, 8, 8, 256)         1024      
_________________________________________________________________
activation_5 (Activation)    (None, 8, 8, 256)         0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 8, 8, 512)         1180160   
_________________________________________________________________
batch_normalization_6 (Batch (None, 8, 8, 512)         2048      
_________________________________________________________________
activation_6 (Activation)    (None, 8, 8, 512)         0         
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 4, 4, 512)         0         
_________________________________________________________________
flatten (Flatten)            (None, 8192)              0         
_________________________________________________________________
dropout (Dropout)            (None, 8192)              0         
_________________________________________________________________
dense (Dense)                (None, 300)               2457900   
_________________________________________________________________
dropout_1 (Dropout)          (None, 300)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 100)               30100     
_________________________________________________________________
dropout_2 (Dropout)          (None, 100)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 10)                1010      
=================================================================
Total params: 4,800,754
Trainable params: 4,798,002
Non-trainable params: 2,752
_________________________________________________________________


model.compile(optimizer=Adam(0.001), loss="categorical_crossentropy", metrics=["accuracy"])


checkpoint_cb = ModelCheckpoint("my_keras_model.h5", save_best_only=True, verbose=1)
early_stopping_cb = EarlyStopping(patience=10, restore_best_weights=True)
reducelr_cb = ReduceLROnPlateau(monitor="val_loss", factor=0.2, patience=5, mode="min", verbose=1)


batch_size = 32
history = model.fit(x=train_images, y=train_oh_labels, batch_size=batch_size, epochs=40, validation_split=0.2, 
                    callbacks=[checkpoint_cb, early_stopping_cb, reducelr_cb])

Epoch 1/40
1250/1250 [==============================] - 44s 34ms/step - loss: 1.9238 - accuracy: 0.2819 - val_loss: 2.1319 - val_accuracy: 0.2787

Epoch 00001: val_loss improved from inf to 2.13189, saving model to my_keras_model.h5
Epoch 2/40
1250/1250 [==============================] - 42s 33ms/step - loss: 1.4577 - accuracy: 0.4533 - val_loss: 1.5571 - val_accuracy: 0.4819

Epoch 00002: val_loss improved from 2.13189 to 1.55712, saving model to my_keras_model.h5
Epoch 3/40
1250/1250 [==============================] - 42s 33ms/step - loss: 1.2312 - accuracy: 0.5560 - val_loss: 1.3039 - val_accuracy: 0.5688

Epoch 00003: val_loss improved from 1.55712 to 1.30385, saving model to my_keras_model.h5
Epoch 4/40
1250/1250 [==============================] - 42s 33ms/step - loss: 1.0752 - accuracy: 0.6200 - val_loss: 1.0629 - val_accuracy: 0.6163

Epoch 00004: val_loss improved from 1.30385 to 1.06294, saving model to my_keras_model.h5
Epoch 5/40
1250/1250 [==============================] - 42s 33ms/step - loss: 0.9580 - accuracy: 0.6638 - val_loss: 1.1420 - val_accuracy: 0.6305

Epoch 00005: val_loss did not improve from 1.06294
Epoch 6/40
1250/1250 [==============================] - 42s 33ms/step - loss: 0.8573 - accuracy: 0.7063 - val_loss: 1.1030 - val_accuracy: 0.6318

Epoch 00006: val_loss did not improve from 1.06294
Epoch 7/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.7809 - accuracy: 0.7356 - val_loss: 0.8965 - val_accuracy: 0.7063

Epoch 00007: val_loss improved from 1.06294 to 0.89652, saving model to my_keras_model.h5
Epoch 8/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.7020 - accuracy: 0.7653 - val_loss: 0.7562 - val_accuracy: 0.7421

Epoch 00008: val_loss improved from 0.89652 to 0.75623, saving model to my_keras_model.h5
Epoch 9/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.6382 - accuracy: 0.7864 - val_loss: 0.6688 - val_accuracy: 0.7723

Epoch 00009: val_loss improved from 0.75623 to 0.66879, saving model to my_keras_model.h5
Epoch 10/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.5479 - accuracy: 0.8186 - val_loss: 0.6772 - val_accuracy: 0.7803

Epoch 00010: val_loss did not improve from 0.66879
Epoch 11/40
1250/1250 [==============================] - 42s 33ms/step - loss: 0.4877 - accuracy: 0.8409 - val_loss: 0.5531 - val_accuracy: 0.8164

Epoch 00011: val_loss improved from 0.66879 to 0.55311, saving model to my_keras_model.h5
Epoch 12/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.4299 - accuracy: 0.8594 - val_loss: 0.5521 - val_accuracy: 0.8094

Epoch 00012: val_loss improved from 0.55311 to 0.55215, saving model to my_keras_model.h5
Epoch 13/40
1250/1250 [==============================] - 44s 35ms/step - loss: 0.3827 - accuracy: 0.8762 - val_loss: 0.6624 - val_accuracy: 0.7792

Epoch 00013: val_loss did not improve from 0.55215
Epoch 14/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.3354 - accuracy: 0.8922 - val_loss: 0.5949 - val_accuracy: 0.8216

Epoch 00014: val_loss did not improve from 0.55215
Epoch 15/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.2998 - accuracy: 0.9046 - val_loss: 0.6030 - val_accuracy: 0.8186

Epoch 00015: val_loss did not improve from 0.55215
Epoch 16/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.2621 - accuracy: 0.9154 - val_loss: 0.7687 - val_accuracy: 0.7766

Epoch 00016: val_loss did not improve from 0.55215
Epoch 17/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.2395 - accuracy: 0.9219 - val_loss: 0.5546 - val_accuracy: 0.8413

Epoch 00017: val_loss did not improve from 0.55215

Epoch 00017: ReduceLROnPlateau reducing learning rate to 0.00020000000949949026.
Epoch 18/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.1311 - accuracy: 0.9587 - val_loss: 0.5208 - val_accuracy: 0.8661

Epoch 00018: val_loss improved from 0.55215 to 0.52075, saving model to my_keras_model.h5
Epoch 19/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0919 - accuracy: 0.9697 - val_loss: 0.5351 - val_accuracy: 0.8759

Epoch 00019: val_loss did not improve from 0.52075
Epoch 20/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0761 - accuracy: 0.9754 - val_loss: 0.5578 - val_accuracy: 0.8733

Epoch 00020: val_loss did not improve from 0.52075
Epoch 21/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0622 - accuracy: 0.9803 - val_loss: 0.5912 - val_accuracy: 0.8744

Epoch 00021: val_loss did not improve from 0.52075
Epoch 22/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0552 - accuracy: 0.9827 - val_loss: 0.6486 - val_accuracy: 0.8708

Epoch 00022: val_loss did not improve from 0.52075
Epoch 23/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0476 - accuracy: 0.9852 - val_loss: 0.6604 - val_accuracy: 0.8714

Epoch 00023: val_loss did not improve from 0.52075

Epoch 00023: ReduceLROnPlateau reducing learning rate to 4.0000001899898055e-05.
Epoch 24/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0346 - accuracy: 0.9891 - val_loss: 0.6614 - val_accuracy: 0.8799

Epoch 00024: val_loss did not improve from 0.52075
Epoch 25/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0298 - accuracy: 0.9908 - val_loss: 0.6705 - val_accuracy: 0.8812

Epoch 00025: val_loss did not improve from 0.52075
Epoch 26/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0266 - accuracy: 0.9915 - val_loss: 0.6950 - val_accuracy: 0.8792

Epoch 00026: val_loss did not improve from 0.52075
Epoch 27/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0232 - accuracy: 0.9926 - val_loss: 0.7273 - val_accuracy: 0.8790

Epoch 00027: val_loss did not improve from 0.52075
Epoch 28/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0214 - accuracy: 0.9931 - val_loss: 0.7615 - val_accuracy: 0.8766

Epoch 00028: val_loss did not improve from 0.52075

Epoch 00028: ReduceLROnPlateau reducing learning rate to 8.000000525498762e-06.


plt.plot(history.history["accuracy"], label="train")
plt.plot(history.history["val_accuracy"], label="valid")
plt.legend()

<matplotlib.legend.Legend at 0x7f30c5e2e310>


model.evaluate(test_images, test_oh_labels, batch_size=batch_size, verbose=1)

313/313 [==============================] - 4s 12ms/step - loss: 0.5660 - accuracy: 0.8602

[0.5659989714622498, 0.8601999878883362]


def build_extended_model():
    tf.keras.backend.clear_session()

    input_tensor = Input(shape=(32, 32, 3))
    x = Conv2D(filters=32, kernel_size=(3, 3), strides=1, padding="same")(input_tensor)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=64, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)

    x = Conv2D(filters=128, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=128, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)
                    
    x = Conv2D(filters=256, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = Conv2D(filters=256, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    # x = MaxPooling2D(pool_size=(2, 2))(x)                 

    x = Conv2D(filters=512, kernel_size=(3, 3), strides=1, padding="same")(x)
    x = BatchNormalization()(x)
    x = Activation("relu")(x)
    x = MaxPooling2D(pool_size=(2, 2))(x)  

    # x = Flatten()(x)      
    x = GlobalAveragePooling2D()(x)
    x = Dropout(rate=0.5)(x)                 
    x = Dense(300, activation="relu")(x)
    x = Dropout(rate=0.3)(x)
    x = Dense(100, activation="relu")(x)
    x = Dropout(rate=0.3)(x)
    output = Dense(10, activation="softmax")(x)

    model = Model(inputs=input_tensor, outputs=output)
    return model


model = build_extended_model()
model.summary()

Model: "model"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 32, 32, 3)]       0         
_________________________________________________________________
conv2d (Conv2D)              (None, 32, 32, 32)        896       
_________________________________________________________________
batch_normalization (BatchNo (None, 32, 32, 32)        128       
_________________________________________________________________
activation (Activation)      (None, 32, 32, 32)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 32, 32, 64)        18496     
_________________________________________________________________
batch_normalization_1 (Batch (None, 32, 32, 64)        256       
_________________________________________________________________
activation_1 (Activation)    (None, 32, 32, 64)        0         
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 16, 16, 64)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 16, 16, 128)       73856     
_________________________________________________________________
batch_normalization_2 (Batch (None, 16, 16, 128)       512       
_________________________________________________________________
activation_2 (Activation)    (None, 16, 16, 128)       0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 16, 16, 128)       147584    
_________________________________________________________________
batch_normalization_3 (Batch (None, 16, 16, 128)       512       
_________________________________________________________________
activation_3 (Activation)    (None, 16, 16, 128)       0         
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 8, 8, 128)         0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 8, 8, 256)         295168    
_________________________________________________________________
batch_normalization_4 (Batch (None, 8, 8, 256)         1024      
_________________________________________________________________
activation_4 (Activation)    (None, 8, 8, 256)         0         
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 8, 8, 256)         590080    
_________________________________________________________________
batch_normalization_5 (Batch (None, 8, 8, 256)         1024      
_________________________________________________________________
activation_5 (Activation)    (None, 8, 8, 256)         0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 8, 8, 512)         1180160   
_________________________________________________________________
batch_normalization_6 (Batch (None, 8, 8, 512)         2048      
_________________________________________________________________
activation_6 (Activation)    (None, 8, 8, 512)         0         
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 4, 4, 512)         0         
_________________________________________________________________
global_average_pooling2d (Gl (None, 512)               0         
_________________________________________________________________
dropout (Dropout)            (None, 512)               0         
_________________________________________________________________
dense (Dense)                (None, 300)               153900    
_________________________________________________________________
dropout_1 (Dropout)          (None, 300)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 100)               30100     
_________________________________________________________________
dropout_2 (Dropout)          (None, 100)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 10)                1010      
=================================================================
Total params: 2,496,754
Trainable params: 2,494,002
Non-trainable params: 2,752
_________________________________________________________________


model.compile(optimizer=Adam(0.001), loss="categorical_crossentropy", metrics=["accuracy"])


checkpoint_cb = ModelCheckpoint("my_keras_model.h5", save_best_only=True, verbose=1)
early_stopping_cb = EarlyStopping(patience=10, restore_best_weights=True)
reducelr_cb = ReduceLROnPlateau(monitor="val_loss", factor=0.2, patience=5, mode="min", verbose=1)


batch_size = 32
history = model.fit(x=train_images, y=train_oh_labels, batch_size=batch_size, epochs=40, validation_split=0.2, 
                    callbacks=[checkpoint_cb, early_stopping_cb, reducelr_cb])

Epoch 1/40
1250/1250 [==============================] - 43s 33ms/step - loss: 1.6409 - accuracy: 0.3811 - val_loss: 1.8103 - val_accuracy: 0.3973

Epoch 00001: val_loss improved from inf to 1.81033, saving model to my_keras_model.h5
Epoch 2/40
1250/1250 [==============================] - 41s 32ms/step - loss: 1.2078 - accuracy: 0.5730 - val_loss: 1.2468 - val_accuracy: 0.5722

Epoch 00002: val_loss improved from 1.81033 to 1.24684, saving model to my_keras_model.h5
Epoch 3/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.9970 - accuracy: 0.6583 - val_loss: 1.3840 - val_accuracy: 0.5496

Epoch 00003: val_loss did not improve from 1.24684
Epoch 4/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.8592 - accuracy: 0.7062 - val_loss: 1.0350 - val_accuracy: 0.6700

Epoch 00004: val_loss improved from 1.24684 to 1.03497, saving model to my_keras_model.h5
Epoch 5/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.7464 - accuracy: 0.7508 - val_loss: 0.8347 - val_accuracy: 0.7101

Epoch 00005: val_loss improved from 1.03497 to 0.83466, saving model to my_keras_model.h5
Epoch 6/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.6532 - accuracy: 0.7872 - val_loss: 1.2667 - val_accuracy: 0.6429

Epoch 00006: val_loss did not improve from 0.83466
Epoch 7/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.5674 - accuracy: 0.8133 - val_loss: 0.7381 - val_accuracy: 0.7651

Epoch 00007: val_loss improved from 0.83466 to 0.73809, saving model to my_keras_model.h5
Epoch 8/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.5080 - accuracy: 0.8344 - val_loss: 0.8084 - val_accuracy: 0.7535

Epoch 00008: val_loss did not improve from 0.73809
Epoch 9/40
1250/1250 [==============================] - 43s 34ms/step - loss: 0.4401 - accuracy: 0.8574 - val_loss: 0.7177 - val_accuracy: 0.7732

Epoch 00009: val_loss improved from 0.73809 to 0.71768, saving model to my_keras_model.h5
Epoch 10/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.3880 - accuracy: 0.8753 - val_loss: 0.6913 - val_accuracy: 0.7834

Epoch 00010: val_loss improved from 0.71768 to 0.69131, saving model to my_keras_model.h5
Epoch 11/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.3414 - accuracy: 0.8883 - val_loss: 0.6659 - val_accuracy: 0.7917

Epoch 00011: val_loss improved from 0.69131 to 0.66594, saving model to my_keras_model.h5
Epoch 12/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.3019 - accuracy: 0.9043 - val_loss: 0.6972 - val_accuracy: 0.8058

Epoch 00012: val_loss did not improve from 0.66594
Epoch 13/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.2722 - accuracy: 0.9141 - val_loss: 0.7045 - val_accuracy: 0.7858

Epoch 00013: val_loss did not improve from 0.66594
Epoch 14/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.2307 - accuracy: 0.9262 - val_loss: 0.7205 - val_accuracy: 0.8036

Epoch 00014: val_loss did not improve from 0.66594
Epoch 15/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.2091 - accuracy: 0.9334 - val_loss: 0.6978 - val_accuracy: 0.8194

Epoch 00015: val_loss did not improve from 0.66594
Epoch 16/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.1835 - accuracy: 0.9427 - val_loss: 0.5686 - val_accuracy: 0.8451

Epoch 00016: val_loss improved from 0.66594 to 0.56863, saving model to my_keras_model.h5
Epoch 17/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.1639 - accuracy: 0.9481 - val_loss: 0.7006 - val_accuracy: 0.8084

Epoch 00017: val_loss did not improve from 0.56863
Epoch 18/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.1541 - accuracy: 0.9513 - val_loss: 0.9184 - val_accuracy: 0.7837

Epoch 00018: val_loss did not improve from 0.56863
Epoch 19/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.1323 - accuracy: 0.9582 - val_loss: 1.0587 - val_accuracy: 0.7794

Epoch 00019: val_loss did not improve from 0.56863
Epoch 20/40
1250/1250 [==============================] - 41s 33ms/step - loss: 0.1216 - accuracy: 0.9617 - val_loss: 0.6943 - val_accuracy: 0.8257

Epoch 00020: val_loss did not improve from 0.56863
Epoch 21/40
1250/1250 [==============================] - 43s 34ms/step - loss: 0.1113 - accuracy: 0.9654 - val_loss: 0.7083 - val_accuracy: 0.8313

Epoch 00021: val_loss did not improve from 0.56863

Epoch 00021: ReduceLROnPlateau reducing learning rate to 0.00020000000949949026.
Epoch 22/40
1250/1250 [==============================] - 40s 32ms/step - loss: 0.0462 - accuracy: 0.9858 - val_loss: 0.6361 - val_accuracy: 0.8707

Epoch 00022: val_loss did not improve from 0.56863
Epoch 23/40
1250/1250 [==============================] - 40s 32ms/step - loss: 0.0215 - accuracy: 0.9934 - val_loss: 0.7566 - val_accuracy: 0.8693

Epoch 00023: val_loss did not improve from 0.56863
Epoch 24/40
1250/1250 [==============================] - 40s 32ms/step - loss: 0.0201 - accuracy: 0.9937 - val_loss: 0.8070 - val_accuracy: 0.8692

Epoch 00024: val_loss did not improve from 0.56863
Epoch 25/40
1250/1250 [==============================] - 42s 34ms/step - loss: 0.0174 - accuracy: 0.9951 - val_loss: 0.8913 - val_accuracy: 0.8655

Epoch 00025: val_loss did not improve from 0.56863
Epoch 26/40
1250/1250 [==============================] - 40s 32ms/step - loss: 0.0182 - accuracy: 0.9950 - val_loss: 0.8484 - val_accuracy: 0.8674

Epoch 00026: val_loss did not improve from 0.56863

Epoch 00026: ReduceLROnPlateau reducing learning rate to 4.0000001899898055e-05.


plt.plot(history.history["accuracy"], label="train")
plt.plot(history.history["val_accuracy"], label="valid")
plt.legend()

<matplotlib.legend.Legend at 0x7f31005c1750>


model.evaluate(test_images, test_oh_labels, batch_size=batch_size, verbose=1)

313/313 [==============================] - 4s 12ms/step - loss: 0.5968 - accuracy: 0.8340

[0.5967729091644287, 0.8339999914169312]

Deep Learning(0913_day7) - CIFAR10 성능올리기

CIFAR10 Dataset 신경망으로 훈련¶

imports¶

데이터 적재¶

데이터 시각화¶

데이터 전처리¶

모델 생성¶

모델 컴파일¶

모델 학습¶

테스트 데이터에서 검증¶

Callback 적용(ModelCheckPoint, EarlyStopping, ReduceLROnPlateau)¶

모델 생성 및 컴파일¶

Callback 정의¶

필터수, 은닉층 수 추가¶

모델 생성 및 컴파일¶

Callback 정의¶

모델 학습¶

테스트 데이터에 대해 평가¶

Global Average Pooling 적용¶

모델 생성 및 컴파일¶

Callback 정의¶

모델 학습¶

테스트 데이터에 대해 평가¶