Python 学習時にデータの一部しか認識されない

前提・実現したいこと

１４４６枚のデータを使って学習を行いたいのですが、学習時のステータスバーを確認すると、データが１２１枚しか使用されていないように見受けられます。
なぜ１４４６枚のデータではなく、１２１枚しか学習に使われていないのか、お分かりの方がいらっしゃいましたらご教示いただけますでしょうか。
お手数をお掛けしますがどうぞよろしくお願いいたします。

発生している問題・エラーメッセージ

Epoch 1/100
121/121 [==============================] - 33s 270ms/step - loss: 0.4005 - accuracy: 0.8859 - val_loss: 1.3700 - val_accuracy: 0.4647

該当のソースコード

Python
1rom keras.models import Model
2from keras.layers import Dense, GlobalAveragePooling2D,Input,Dropout,Activation
3from keras.applications.resnet50 import ResNet50
4from keras.preprocessing.image import ImageDataGenerator
5from keras.optimizers import Adam
6from keras.callbacks import CSVLogger,EarlyStopping
7import numpy as np
8from keras import backend as K
9from keras.engine.topology import Layer
10import numpy as np
11import tensorflow as tf
12from keras.preprocessing.image import load_img, img_to_array, array_to_img
13import os
14import glob
15import cv2
16from PIL import Image
17import matplotlib.pyplot as plt
18import keras
19from keras import callbacks
20from keras.utils import to_categorical
21%matplotlib inline
22
23os.environ["CUDA_VISIBLE_DEVICES"]="0"#specify GPU
24
25image_size_width, image_size_height = (299, 299)
26
27folder = ["COVID-19_resize_train_1_2", "Viral Pneumonia_resize_train_1_2", "NORMAL_resize_train_1_2"]
28x_train = []
29y_train = []
30for index, name in enumerate(folder):
31    dir = "./" + name
32    files = glob.glob(dir + "/*.png")
33    for i, file in enumerate(files):
34        image = Image.open(file)
35        image = image.convert("RGB")
36        image = image.resize((image_size_width, image_size_height))
37        data = np.asarray(image)
38        x_train.append(data)
39        y_train.append(index)
40
41x_train = np.array(x_train)
42y_train = np.array(y_train)
43print('x_train.shape', x_train.shape)
44print('y_train.shape', y_train.shape)
45#x_train.shape (1446, 299, 299, 3)
46#y_train.shape (1446,)
47
48folder = ["COVID-19_resize_train_3", "Viral Pneumonia_resize_train_3", "NORMAL_resize_train_3"]
49x_test = []
50y_test = []
51for index, name in enumerate(folder):
52    dir = "./" + name
53    files = glob.glob(dir + "/*.png")
54    for i, file in enumerate(files):
55        image = Image.open(file)
56        image = image.convert("RGB")
57        image = image.resize((image_size_width, image_size_height))
58        data = np.asarray(image)
59        x_test.append(data)
60        y_test.append(index)
61
62x_test = np.array(x_test)
63y_test = np.array(y_test)
64print('x_test.shape', x_test.shape)
65print('y_test.shape', y_test.shape)
66#x_test.shape (723, 299, 299, 3)
67#y_test.shape (723,)
68
69input_shape = (299, 299, 3)
70classes = 3
71batchsize = 12
72epochs=100
73
74
75base_model=keras.applications.densenet.DenseNet201(input_shape=input_shape,
76                                                   weights='imagenet',
77                                                   include_top=False)
78    
79c = base_model.output
80c = GlobalAveragePooling2D()(c)
81c = keras.layers.Lambda(lambda xx: 5*(xx)/K.sqrt(K.sum(xx**2)))(c) #metric learning
82c = Dense(classes, activation='softmax')(c)
83model = Model(inputs=base_model.input,outputs=c)
84
85model.summary()
86
87model.compile(loss='categorical_crossentropy',
88                  optimizer=Adam(lr=0.001),
89                  metrics=['accuracy'])
90    
91callbacks_list = [
92        callbacks.ModelCheckpoint(
93        filepath="model.DenseNet201.ep{epoch:02d}.h5",
94        save_weights_only=False,
95        save_best_only=True)]#False)]
96    
97history = model.fit(x_train, to_categorical(y_train, classes), batch_size=batchsize, epochs=epochs, verbose=1, validation_data=(x_test, to_categorical(y_test, classes)), 
98callbacks=callbacks_list)
99