第T4周：猴痘病识别

最新推荐文章于 2024-07-25 18:55:58 发布

SHiHy^^

最新推荐文章于 2024-07-25 18:55:58 发布

阅读量902

点赞数 18

文章标签： python 分类神经网络 tensorflow2

本文链接：https://blog.csdn.net/weixin_62819126/article/details/137210285

版权

>- **🍨 本文为[🔗365天深度学习训练营](小团体~第六波) 中的学习记录博客**
>- **🍖 原作者：[K同学啊 | 接辅导、项目定制](K同学啊-CSDN博客)**

🏡 我的环境：

语言环境：Python3.7

编译器：jupyter lab

深度学习环境：Pytorch

一、前期工作

1. 设置GPU

如果使用的是CPU可以忽略这步

from tensorflow       import keras
from tensorflow.keras import layers,models
import os, PIL, pathlib
import matplotlib.pyplot as plt
import tensorflow as tf

gpus = tf.config.list_physical_devices("GPU")

if gpus:
    gpu0 = gpus[0]                                        #如果有多个GPU，仅使用第0个GPU
    tf.config.experimental.set_memory_growth(gpu0, True)  #设置GPU显存用量按需使用
    tf.config.set_visible_devices([gpu0],"GPU")
    
gpus

2. 导入数据

data_dir = "./dataset/t4/"

data_dir = pathlib.Path(data_dir)

3. 查看数据

image_count = len(list(data_dir.glob('*/*.jpg')))
print("图片总数为",image_count)

图片总数为 2142

Monkeypox = list(data_dir.glob("Monkeypox/*.jpg"))
PIL.Image.open(str(Monkeypox[0]))

二、数据预处理

1. 加载数据

使用image_dataset_from_directory方法将磁盘中的数据加载到tf.data.Dataset中

测试集与验证集的关系：

验证集并没有参与训练过程梯度下降过程的，狭义上来讲是没有参与模型的参数训练更新的。
但是广义上来讲，验证集存在的意义确实参与了一个“人工调参”的过程，我们根据每一个epoch训练之后模型在valid data上的表现来决定是否需要训练进行early stop，或者根据这个过程模型的性能变化来调整模型的超参数，如学习率，batch_size等等。
因此，我们也可以认为，验证集也参与了训练，但是并没有使得模型去overfit验证集

batch_size = 32
img_height = 224
img_width = 224

train_ds = tf.keras.preprocessing.image_dataset_from_directory(
    data_dir,
    validation_split=0.2,
    subset="training",
    seed=123,
    image_size=(img_height,img_width),
    batch_size=batch_size)

Found 2142 files belonging to 2 classes.
Using 428 files for validation.

val_ds = tf.keras.preprocessing.image_dataset_from_directory(
    data_dir,
    validation_split=0.2,
    subset="validation",
    seed=123,
    image_size=(img_height,img_width),
    batch_size=batch_size)

Found 2142 files belonging to 2 classes.
Using 428 files for validation.

class_names = train_ds.class_names
class_names

['Monkeypox', 'Others']

2. 可视化数据

plt.figure(figsize=(20,10))

for images, labels in train_ds.take(1):
    for i in range(20):
        ax = plt.subplot(5,10,i+1)
        
        plt.imshow(images[i].numpy().astype("uint8"))
        plt.title(class_names[labels[i]])
        
        plt.axis("off")

3. 再次检查数据

for image_batch, labels_batch in train_ds:
    print(image_batch.shape)
    print(labels_batch.shape)
    break

(32, 224, 224, 3)
(32,)

Image_batch是形状的张量（32,224,224,3）。这是一批形状224x224x3的32张图片（最后一维指的是彩色通道RGB）。

Label_batch是形状（32，）的张量，这些标签对应32张图片

4. 配置数据集

shuffle() ：打乱数据，关于此函数的详细介绍可以参考：数据集shuffle方法中buffer_size的理解 - 知乎

prefetch() ：预取数据，加速运行

prefetch()功能详细介绍：CPU 正在准备数据时，加速器处于空闲状态。相反，当加速器正在训练模型时，CPU 处于空闲状态。因此，训练所用的时间是 CPU 预处理时间和加速器训练时间的总和。prefetch()将训练步骤的预处理和模型执行过程重叠到一起。当加速器正在执行第 N 个训练步时，CPU 正在准备第 N+1 步的数据。这样做不仅可以最大限度地缩短训练的单步用时（而不是总用时），而且可以缩短提取和转换数据所需的时间。如果不使用prefetch()，CPU 和 GPU/TPU 在大部分时间都处于空闲状态

使用prefetch()可显著减少空闲时间

cache() ：将数据集缓存到内存当中，加速运行

AUTOTUNE = tf.data.experimental.AUTOTUNE

train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)
val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)

三、构建CNN网络

卷积神经网络（CNN）的输入是张量 (Tensor) 形式的 (image_height, image_width, color_channels)，包含了图像高度、宽度及颜色信息。不需要输入batch size。color_channels 为 (R,G,B) 分别对应 RGB 的三个颜色通道（color channel）。在此示例中，我们的 CNN 输入的形状是 (224, 224, 3)即彩色图像。我们需要在声明第一层时将形状赋值给参数input_shape。

num_classes = 2

model = models.Sequential([
    layers.experimental.preprocessing.Rescaling(1./255,input_shape=(img_height, img_width, 3)),
    layers.Conv2D(16,(3,3),activation='relu',input_shape=(img_height,img_width,3)),
    layers.AveragePooling2D((2,2)),
    layers.Conv2D(32,(3,3),activation='relu'),
    layers.AveragePooling2D((2,2)),
    layers.Dropout(0.3),
    layers.Conv2D(64,(3,3),activation='relu'),
    layers.Dropout(0.3),
    
    layers.Flatten(),
    layers.Dense(128,activation='relu'),
    layers.Dense(num_classes)
])

model.summary()

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
rescaling_3 (Rescaling)      (None, 224, 224, 3)       0         
_________________________________________________________________
conv2d_9 (Conv2D)            (None, 222, 222, 16)      448       
_________________________________________________________________
average_pooling2d_6 (Average (None, 111, 111, 16)      0         
_________________________________________________________________
conv2d_10 (Conv2D)           (None, 109, 109, 32)      4640      
_________________________________________________________________
average_pooling2d_7 (Average (None, 54, 54, 32)        0         
_________________________________________________________________
dropout_6 (Dropout)          (None, 54, 54, 32)        0         
_________________________________________________________________
conv2d_11 (Conv2D)           (None, 52, 52, 64)        18496     
_________________________________________________________________
dropout_7 (Dropout)          (None, 52, 52, 64)        0         
_________________________________________________________________
flatten_3 (Flatten)          (None, 173056)            0         
_________________________________________________________________
dense_4 (Dense)              (None, 128)               22151296  
_________________________________________________________________
dense_5 (Dense)              (None, 2)                 258       
=================================================================
Total params: 22,175,138
Trainable params: 22,175,138
Non-trainable params: 0
_________________________________________________________________

四、编译

在准备对模型进行训练之前，还需要再对其进行一些设置。以下内容是在模型的编译步骤中添加的：

损失函数（loss）：用于衡量模型在训练期间的准确率。

优化器（optimizer）：决定模型如何根据其看到的数据和自身的损失函数进行更新。

指标（metrics）：用于监控训练和测试步骤。以下示例使用了准确率，即被正确分类的图像的比率。

opt = tf.keras.optimizers.Adam(learning_rate=1e-4)

model.compile(optimizer=opt,
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

五、训练模型

关于ModelCheckpoint的详细介绍可参考文章 🔗ModelCheckpoint 讲解【TensorFlow2入门手册】

from tensorflow.keras.callbacks import ModelCheckpoint

epochs = 50

checkpointer = ModelCheckpoint('best_model.h5',
                              monitor='val_accuracy',
                              verbose=1,
                              save_best_only=True,
                              save_weights_only=True)

history = model.fit(train_ds,
                   validation_data=val_ds,
                   epochs=epochs,
                   callbacks=[checkpointer])

Epoch 1/50
54/54 [==============================] - ETA: 0s - loss: 0.7364 - accuracy: 0.5245
Epoch 00001: val_accuracy improved from -inf to 0.54206, saving model to best_model.h5
54/54 [==============================] - 67s 1s/step - loss: 0.7364 - accuracy: 0.5245 - val_loss: 0.6729 - val_accuracy: 0.5421
Epoch 2/50
54/54 [==============================] - ETA: 0s - loss: 0.6652 - accuracy: 0.6050
Epoch 00002: val_accuracy improved from 0.54206 to 0.61449, saving model to best_model.h5
54/54 [==============================] - 51s 951ms/step - loss: 0.6652 - accuracy: 0.6050 - val_loss: 0.6493 - val_accuracy: 0.6145
Epoch 3/50
54/54 [==============================] - ETA: 0s - loss: 0.6341 - accuracy: 0.6389
Epoch 00003: val_accuracy improved from 0.61449 to 0.66121, saving model to best_model.h5
54/54 [==============================] - 52s 956ms/step - loss: 0.6341 - accuracy: 0.6389 - val_loss: 0.6114 - val_accuracy: 0.6612
Epoch 4/50
54/54 [==============================] - ETA: 0s - loss: 0.6396 - accuracy: 0.6295
Epoch 00004: val_accuracy did not improve from 0.66121
54/54 [==============================] - 48s 883ms/step - loss: 0.6396 - accuracy: 0.6295 - val_loss: 0.6311 - val_accuracy: 0.6402
Epoch 5/50
54/54 [==============================] - ETA: 0s - loss: 0.5984 - accuracy: 0.6733
Epoch 00005: val_accuracy improved from 0.66121 to 0.70561, saving model to best_model.h5
54/54 [==============================] - 58s 1s/step - loss: 0.5984 - accuracy: 0.6733 - val_loss: 0.5847 - val_accuracy: 0.7056
Epoch 6/50
54/54 [==============================] - ETA: 0s - loss: 0.5770 - accuracy: 0.6914
Epoch 00006: val_accuracy improved from 0.70561 to 0.70794, saving model to best_model.h5
54/54 [==============================] - 52s 970ms/step - loss: 0.5770 - accuracy: 0.6914 - val_loss: 0.5703 - val_accuracy: 0.7079
Epoch 7/50
54/54 [==============================] - ETA: 0s - loss: 0.5253 - accuracy: 0.7427
Epoch 00007: val_accuracy did not improve from 0.70794
54/54 [==============================] - 56s 1s/step - loss: 0.5253 - accuracy: 0.7427 - val_loss: 0.5814 - val_accuracy: 0.7056
Epoch 8/50
54/54 [==============================] - ETA: 0s - loss: 0.4845 - accuracy: 0.7643
Epoch 00008: val_accuracy improved from 0.70794 to 0.73364, saving model to best_model.h5
54/54 [==============================] - 53s 973ms/step - loss: 0.4845 - accuracy: 0.7643 - val_loss: 0.5168 - val_accuracy: 0.7336
Epoch 9/50
54/54 [==============================] - ETA: 0s - loss: 0.4461 - accuracy: 0.7853
Epoch 00009: val_accuracy improved from 0.73364 to 0.76402, saving model to best_model.h5
54/54 [==============================] - 59s 1s/step - loss: 0.4461 - accuracy: 0.7853 - val_loss: 0.4640 - val_accuracy: 0.7640
Epoch 10/50
54/54 [==============================] - ETA: 0s - loss: 0.4262 - accuracy: 0.8162
Epoch 00010: val_accuracy did not improve from 0.76402
54/54 [==============================] - 51s 949ms/step - loss: 0.4262 - accuracy: 0.8162 - val_loss: 0.5703 - val_accuracy: 0.7173
Epoch 11/50
54/54 [==============================] - ETA: 0s - loss: 0.4144 - accuracy: 0.8063
Epoch 00011: val_accuracy did not improve from 0.76402
54/54 [==============================] - 58s 1s/step - loss: 0.4144 - accuracy: 0.8063 - val_loss: 0.4668 - val_accuracy: 0.7617
Epoch 12/50
54/54 [==============================] - ETA: 0s - loss: 0.3724 - accuracy: 0.8372
Epoch 00012: val_accuracy improved from 0.76402 to 0.80140, saving model to best_model.h5
54/54 [==============================] - 54s 1s/step - loss: 0.3724 - accuracy: 0.8372 - val_loss: 0.4342 - val_accuracy: 0.8014
Epoch 13/50
54/54 [==============================] - ETA: 0s - loss: 0.3384 - accuracy: 0.8489
Epoch 00013: val_accuracy improved from 0.80140 to 0.82710, saving model to best_model.h5
54/54 [==============================] - 55s 1s/step - loss: 0.3384 - accuracy: 0.8489 - val_loss: 0.4052 - val_accuracy: 0.8271
Epoch 14/50
54/54 [==============================] - ETA: 0s - loss: 0.3083 - accuracy: 0.8716
Epoch 00014: val_accuracy improved from 0.82710 to 0.83879, saving model to best_model.h5
54/54 [==============================] - 54s 1s/step - loss: 0.3083 - accuracy: 0.8716 - val_loss: 0.3880 - val_accuracy: 0.8388
Epoch 15/50
54/54 [==============================] - ETA: 0s - loss: 0.3042 - accuracy: 0.8839
Epoch 00015: val_accuracy improved from 0.83879 to 0.86215, saving model to best_model.h5
54/54 [==============================] - 52s 971ms/step - loss: 0.3042 - accuracy: 0.8839 - val_loss: 0.3893 - val_accuracy: 0.8621
Epoch 16/50
54/54 [==============================] - ETA: 0s - loss: 0.3068 - accuracy: 0.8728
Epoch 00016: val_accuracy did not improve from 0.86215
54/54 [==============================] - 53s 975ms/step - loss: 0.3068 - accuracy: 0.8728 - val_loss: 0.4131 - val_accuracy: 0.8248
Epoch 17/50
54/54 [==============================] - ETA: 0s - loss: 0.2609 - accuracy: 0.8950
Epoch 00017: val_accuracy improved from 0.86215 to 0.87383, saving model to best_model.h5
54/54 [==============================] - 52s 970ms/step - loss: 0.2609 - accuracy: 0.8950 - val_loss: 0.3935 - val_accuracy: 0.8738
Epoch 18/50
54/54 [==============================] - ETA: 0s - loss: 0.2490 - accuracy: 0.8991
Epoch 00018: val_accuracy did not improve from 0.87383
54/54 [==============================] - 62s 1s/step - loss: 0.2490 - accuracy: 0.8991 - val_loss: 0.4317 - val_accuracy: 0.8364
Epoch 19/50
54/54 [==============================] - ETA: 0s - loss: 0.2328 - accuracy: 0.9148
Epoch 00019: val_accuracy did not improve from 0.87383
54/54 [==============================] - 64s 1s/step - loss: 0.2328 - accuracy: 0.9148 - val_loss: 0.4172 - val_accuracy: 0.8364
Epoch 20/50
54/54 [==============================] - ETA: 0s - loss: 0.2398 - accuracy: 0.9037
Epoch 00020: val_accuracy did not improve from 0.87383
54/54 [==============================] - 57s 1s/step - loss: 0.2398 - accuracy: 0.9037 - val_loss: 0.3751 - val_accuracy: 0.8598
Epoch 21/50
54/54 [==============================] - ETA: 0s - loss: 0.2258 - accuracy: 0.9096
Epoch 00021: val_accuracy did not improve from 0.87383
54/54 [==============================] - 51s 950ms/step - loss: 0.2258 - accuracy: 0.9096 - val_loss: 0.3765 - val_accuracy: 0.8668
Epoch 22/50
54/54 [==============================] - ETA: 0s - loss: 0.1866 - accuracy: 0.9341
Epoch 00022: val_accuracy did not improve from 0.87383
54/54 [==============================] - 52s 964ms/step - loss: 0.1866 - accuracy: 0.9341 - val_loss: 0.3741 - val_accuracy: 0.8645
Epoch 23/50
54/54 [==============================] - ETA: 0s - loss: 0.1829 - accuracy: 0.9341
Epoch 00023: val_accuracy did not improve from 0.87383
54/54 [==============================] - 52s 954ms/step - loss: 0.1829 - accuracy: 0.9341 - val_loss: 0.3990 - val_accuracy: 0.8551
Epoch 24/50
54/54 [==============================] - ETA: 0s - loss: 0.1767 - accuracy: 0.9306
Epoch 00024: val_accuracy did not improve from 0.87383
54/54 [==============================] - 52s 967ms/step - loss: 0.1767 - accuracy: 0.9306 - val_loss: 0.3802 - val_accuracy: 0.8738
Epoch 25/50
54/54 [==============================] - ETA: 0s - loss: 0.1540 - accuracy: 0.9428
Epoch 00025: val_accuracy improved from 0.87383 to 0.87617, saving model to best_model.h5
54/54 [==============================] - 52s 972ms/step - loss: 0.1540 - accuracy: 0.9428 - val_loss: 0.3867 - val_accuracy: 0.8762
Epoch 26/50
54/54 [==============================] - ETA: 0s - loss: 0.1473 - accuracy: 0.9469
Epoch 00026: val_accuracy did not improve from 0.87617
54/54 [==============================] - 52s 962ms/step - loss: 0.1473 - accuracy: 0.9469 - val_loss: 0.3829 - val_accuracy: 0.8738
Epoch 27/50
54/54 [==============================] - ETA: 0s - loss: 0.1427 - accuracy: 0.9481
Epoch 00027: val_accuracy did not improve from 0.87617
54/54 [==============================] - 51s 944ms/step - loss: 0.1427 - accuracy: 0.9481 - val_loss: 0.4618 - val_accuracy: 0.8435
Epoch 28/50
54/54 [==============================] - ETA: 0s - loss: 0.1282 - accuracy: 0.9539
Epoch 00028: val_accuracy did not improve from 0.87617
54/54 [==============================] - 52s 956ms/step - loss: 0.1282 - accuracy: 0.9539 - val_loss: 0.4185 - val_accuracy: 0.8668
Epoch 29/50
54/54 [==============================] - ETA: 0s - loss: 0.1244 - accuracy: 0.9551
Epoch 00029: val_accuracy did not improve from 0.87617
54/54 [==============================] - 51s 948ms/step - loss: 0.1244 - accuracy: 0.9551 - val_loss: 0.3895 - val_accuracy: 0.8692
Epoch 30/50
54/54 [==============================] - ETA: 0s - loss: 0.1226 - accuracy: 0.9597
Epoch 00030: val_accuracy did not improve from 0.87617
54/54 [==============================] - 52s 957ms/step - loss: 0.1226 - accuracy: 0.9597 - val_loss: 0.3974 - val_accuracy: 0.8715
Epoch 31/50
54/54 [==============================] - ETA: 0s - loss: 0.1094 - accuracy: 0.9644
Epoch 00031: val_accuracy improved from 0.87617 to 0.87850, saving model to best_model.h5
54/54 [==============================] - 53s 985ms/step - loss: 0.1094 - accuracy: 0.9644 - val_loss: 0.4113 - val_accuracy: 0.8785
Epoch 32/50
54/54 [==============================] - ETA: 0s - loss: 0.0973 - accuracy: 0.9737
Epoch 00032: val_accuracy did not improve from 0.87850
54/54 [==============================] - 50s 928ms/step - loss: 0.0973 - accuracy: 0.9737 - val_loss: 0.4070 - val_accuracy: 0.8762
Epoch 33/50
54/54 [==============================] - ETA: 0s - loss: 0.1078 - accuracy: 0.9667
Epoch 00033: val_accuracy did not improve from 0.87850
54/54 [==============================] - 49s 915ms/step - loss: 0.1078 - accuracy: 0.9667 - val_loss: 0.4207 - val_accuracy: 0.8692
Epoch 34/50
54/54 [==============================] - ETA: 0s - loss: 0.0991 - accuracy: 0.9638
Epoch 00034: val_accuracy did not improve from 0.87850
54/54 [==============================] - 51s 942ms/step - loss: 0.0991 - accuracy: 0.9638 - val_loss: 0.4123 - val_accuracy: 0.8621
Epoch 35/50
54/54 [==============================] - ETA: 0s - loss: 0.0820 - accuracy: 0.9755
Epoch 00035: val_accuracy did not improve from 0.87850
54/54 [==============================] - 51s 946ms/step - loss: 0.0820 - accuracy: 0.9755 - val_loss: 0.4469 - val_accuracy: 0.8645
Epoch 36/50
54/54 [==============================] - ETA: 0s - loss: 0.0849 - accuracy: 0.9720
Epoch 00036: val_accuracy did not improve from 0.87850
54/54 [==============================] - 48s 896ms/step - loss: 0.0849 - accuracy: 0.9720 - val_loss: 0.4354 - val_accuracy: 0.8668
Epoch 37/50
54/54 [==============================] - ETA: 0s - loss: 0.0847 - accuracy: 0.9708
Epoch 00037: val_accuracy did not improve from 0.87850
54/54 [==============================] - 49s 917ms/step - loss: 0.0847 - accuracy: 0.9708 - val_loss: 0.4577 - val_accuracy: 0.8575
Epoch 38/50
54/54 [==============================] - ETA: 0s - loss: 0.0720 - accuracy: 0.9813
Epoch 00038: val_accuracy did not improve from 0.87850
54/54 [==============================] - 46s 848ms/step - loss: 0.0720 - accuracy: 0.9813 - val_loss: 0.4645 - val_accuracy: 0.8645
Epoch 39/50
54/54 [==============================] - ETA: 0s - loss: 0.0671 - accuracy: 0.9767
Epoch 00039: val_accuracy did not improve from 0.87850
54/54 [==============================] - 46s 847ms/step - loss: 0.0671 - accuracy: 0.9767 - val_loss: 0.4596 - val_accuracy: 0.8738
Epoch 40/50
54/54 [==============================] - ETA: 0s - loss: 0.0612 - accuracy: 0.9854
Epoch 00040: val_accuracy did not improve from 0.87850
54/54 [==============================] - 46s 844ms/step - loss: 0.0612 - accuracy: 0.9854 - val_loss: 0.4922 - val_accuracy: 0.8692
Epoch 41/50
54/54 [==============================] - ETA: 0s - loss: 0.1007 - accuracy: 0.9632
Epoch 00041: val_accuracy did not improve from 0.87850
54/54 [==============================] - 46s 847ms/step - loss: 0.1007 - accuracy: 0.9632 - val_loss: 0.5131 - val_accuracy: 0.8692
Epoch 42/50
54/54 [==============================] - ETA: 0s - loss: 0.0921 - accuracy: 0.9667
Epoch 00042: val_accuracy did not improve from 0.87850
54/54 [==============================] - 46s 853ms/step - loss: 0.0921 - accuracy: 0.9667 - val_loss: 0.4745 - val_accuracy: 0.8645
Epoch 43/50
54/54 [==============================] - ETA: 0s - loss: 0.0606 - accuracy: 0.9831
Epoch 00043: val_accuracy did not improve from 0.87850
54/54 [==============================] - 46s 843ms/step - loss: 0.0606 - accuracy: 0.9831 - val_loss: 0.4889 - val_accuracy: 0.8785
Epoch 44/50
54/54 [==============================] - ETA: 0s - loss: 0.0648 - accuracy: 0.9825
Epoch 00044: val_accuracy did not improve from 0.87850
54/54 [==============================] - 45s 837ms/step - loss: 0.0648 - accuracy: 0.9825 - val_loss: 0.4670 - val_accuracy: 0.8785
Epoch 45/50
54/54 [==============================] - ETA: 0s - loss: 0.0573 - accuracy: 0.9842
Epoch 00045: val_accuracy improved from 0.87850 to 0.88318, saving model to best_model.h5
54/54 [==============================] - 45s 842ms/step - loss: 0.0573 - accuracy: 0.9842 - val_loss: 0.4668 - val_accuracy: 0.8832
Epoch 46/50
54/54 [==============================] - ETA: 0s - loss: 0.0465 - accuracy: 0.9837
Epoch 00046: val_accuracy did not improve from 0.88318
54/54 [==============================] - 45s 837ms/step - loss: 0.0465 - accuracy: 0.9837 - val_loss: 0.4830 - val_accuracy: 0.8762
Epoch 47/50
54/54 [==============================] - ETA: 0s - loss: 0.0487 - accuracy: 0.9813
Epoch 00047: val_accuracy did not improve from 0.88318
54/54 [==============================] - 45s 832ms/step - loss: 0.0487 - accuracy: 0.9813 - val_loss: 0.5334 - val_accuracy: 0.8645
Epoch 48/50
54/54 [==============================] - ETA: 0s - loss: 0.0505 - accuracy: 0.9895
Epoch 00048: val_accuracy did not improve from 0.88318
54/54 [==============================] - 45s 829ms/step - loss: 0.0505 - accuracy: 0.9895 - val_loss: 0.5237 - val_accuracy: 0.8575
Epoch 49/50
54/54 [==============================] - ETA: 0s - loss: 0.0499 - accuracy: 0.9877
Epoch 00049: val_accuracy did not improve from 0.88318
54/54 [==============================] - 45s 835ms/step - loss: 0.0499 - accuracy: 0.9877 - val_loss: 0.4937 - val_accuracy: 0.8808
Epoch 50/50
54/54 [==============================] - ETA: 0s - loss: 0.0477 - accuracy: 0.9842
Epoch 00050: val_accuracy did not improve from 0.88318
54/54 [==============================] - 45s 837ms/step - loss: 0.0477 - accuracy: 0.9842 - val_loss: 0.5719 - val_accuracy: 0.8715

六、模型评估

1. Loss与Accuracy图

acc = history.history['accuracy']
val_acc = history.history['val_accuracy']

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

epochs_range = range(epochs)

plt.figure(figsize=(12, 4))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, acc, label='Training Accuracy')
plt.plot(epochs_range, val_acc, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, label='Training Loss')
plt.plot(epochs_range, val_loss, label='Validation Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()

2. 指定图片进行预测

# 加载效果最好的模型权重
model.load_weights('best_model.h5')

from PIL import Image
import numpy as np

# img = Image.open("./dataset/t4/Monkeypox/M06_01_04.jpg")  #这里选择你需要预测的图片
img = Image.open("./dataset/t4/Others/NM15_02_11.jpg")  #这里选择你需要预测的图片

# 将图像转换为NumPy数组
img_array = np.array(img)

# 使用TensorFlow的函数将NumPy数组转换为张量
img_tensor = tf.convert_to_tensor(img_array, dtype=tf.float32)
image = tf.image.resize(img_tensor, [img_height, img_width])

img_array = tf.expand_dims(image, 0) 

predictions = model.predict(img_array) # 这里选用你已经训练好的模型
print("预测结果为：",class_names[np.argmax(predictions)])

预测结果为： Others

七、总结

1.ModelCheckpoint 讲解

tf.keras.callbacks.ModelCheckpoint(
    filepath, monitor='val_loss', verbose=0, save_best_only=False,
    save_weights_only=False, mode='auto', save_freq='epoch',
    options=None, **kwargs
)

filepath：字符串，保存模型路径
monitor:需要监视的值
verbose:：信息展示模式，0或1(checkpoint的保存信息，类似Epoch 00001: saving model to …)
save_best_only:当设置为True时，监测值有改进时才会保存当前的模型（ the latest best model according to the quantity monitored will not be overwritten）
mode:‘auto’，‘min’，‘max’之一，在save_best_only=True时决定性能最佳模型的评判准则，例如，当监测值为val_acc时，模式应为max，当监测值为val_loss时，模式应为min。在auto模式下，评价准则由被监测值的名字自动推断。
save_weights_only:若设置为True，则只保存模型权重，否则将保存整个模型（包括模型结构，配置信息等）
period:CheckPoint之间的间隔的epoch数