Tensorflow实现深度学习案例4：猴痘病识别

小叮当爱咖啡

于 2024-08-02 16:02:59 发布

阅读量847

点赞数 16

文章标签：深度学习 tensorflow 人工智能

本文链接：https://blog.csdn.net/2301_76606951/article/details/140871897

版权

本文为为🔗365天深度学习训练营内部文章

原作者：K同学啊

一前期准备

1.导入数据

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

data_dir = "./ills/"

data_dir = pathlib.Path(data_dir)
data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[1] for path in data_paths]
classeNames

['Monkeypox', 'Others']

image_count = len(list(data_dir.glob('*/*.jpg')))

print("图片总数为：",image_count)

图片总数为： 2142

2.查看数据

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

import matplotlib.pyplot as plt
from PIL import Image

# 指定图像文件夹路径
image_folder = './ills/Monkeypox/'

# 获取文件夹中的所有图像文件
image_files = [f for f in os.listdir(image_folder) if f.endswith((".jpg", ".png", ".jpeg"))]

# 创建Matplotlib图像
fig, axes = plt.subplots(3, 8, figsize=(16, 6))

# 使用列表推导式加载和显示图像
for ax, img_file in zip(axes.flat, image_files):
    img_path = os.path.join(image_folder, img_file)
    img = Image.open(img_path)
    ax.imshow(img)
    ax.axis('off')

# 显示图像
plt.tight_layout()
plt.show()

3.数据预处理

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 = 64
img_height = 224
img_width = 224

train_ds = tf.keras.preprocessing.image_dataset_from_directory(
    data_dir,
    validation_split=0.4,
    subset="training",
    seed=123,
    image_size=(img_height, img_width),
    batch_size=batch_size)

val_ds = tf.keras.preprocessing.image_dataset_from_directory(
    data_dir,
    validation_split=0.4,
    subset="validation",
    seed=123,
    image_size=(img_height, img_width),
    batch_size=batch_size)

我们可以通过class_names输出数据集的标签。标签将按字母顺序对应于目录名称

class_names = train_ds.class_names
print(class_names)

['Monkeypox', 'Others']

4.再次查看数据

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")

train_ds.take(1) 从数据集中获取一个批次的数据。每个批次包含 images 和 labels。 images 是一组图像数据，通常是一个张量。 labels 是这些图像的标签，通常是一个张量，包含每个图像的类别编号。

plt.imshow(images[i].numpy().astype("uint8")) images[i] 选取当前批次中的第 i 张图像。 images[i].numpy() 将 TensorFlow 张量转换为 NumPy 数组。 .astype("uint8") 确保图像数据是无符号 8 位整型，这是显示图像时常用的数据类型。 plt.imshow 将图像显示在当前的子图中。 5. plt.title(class_names[labels[i]]) class_names 是一个列表，其中包含所有类别的名称。labels[i] 是当前图像的标签，表示类别的索引。

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

5. 配置数据集

shuffle() ：打乱数据，关于此函数的详细介绍可以参考：https://zhuanlan.zhihu.com/p/42417456

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.AUTOTUNE

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

二、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)), # 卷积层1，卷积核3*3  
    layers.AveragePooling2D((2, 2)),               # 池化层1，2*2采样
    layers.Conv2D(32, (3, 3), activation='relu'),  # 卷积层2，卷积核3*3
    layers.AveragePooling2D((2, 2)),               # 池化层2，2*2采样
    layers.Dropout(0.3),  
    layers.Conv2D(64, (3, 3), activation='relu'),  # 卷积层3，卷积核3*3
    layers.Dropout(0.3),  
    
    layers.Flatten(),                       # Flatten层，连接卷积层与全连接层
    layers.Dense(128, activation='relu'),   # 全连接层，特征进一步提取
    layers.Dense(num_classes)               # 输出层，输出预期结果
])

model.summary()  # 打印网络结构

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 rescaling (Rescaling)       (None, 224, 224, 3)       0         
                                                                 
 conv2d (Conv2D)             (None, 222, 222, 16)      448       
                                                                 
 average_pooling2d (AverageP  (None, 111, 111, 16)     0         
 ooling2D)                                                       
                                                                 
 conv2d_1 (Conv2D)           (None, 109, 109, 32)      4640      
                                                                 
 average_pooling2d_1 (Averag  (None, 54, 54, 32)       0         
 ePooling2D)                                                     
                                                                 
 dropout (Dropout)           (None, 54, 54, 32)        0         
                                                                 
 conv2d_2 (Conv2D)           (None, 52, 52, 64)        18496     
                                                                 
 dropout_1 (Dropout)         (None, 52, 52, 64)        0         
                                                                 
 flatten (Flatten)           (None, 173056)            0         
                                                                 
 dense (Dense)               (None, 128)               22151296  
                                                                 
 dense_1 (Dense)             (None, 2)                 258       
                                                                 
=================================================================
Total params: 22,175,138
Trainable params: 22,175,138
Non-trainable params: 0
_________________________________________________________________

三、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。

# 设置优化器
opt = tf.keras.optimizers.Adam(learning_rate=1e-4)

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

四、训练模型

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
21/21 [==============================] - ETA: 0s - loss: 0.0558 - accuracy: 0.9876
Epoch 1: val_accuracy improved from -inf to 0.93575, saving model to best_model.h5
21/21 [==============================] - 22s 1s/step - loss: 0.0558 - accuracy: 0.9876 - val_loss: 0.2469 - val_accuracy: 0.9357
Epoch 2/50
21/21 [==============================] - ETA: 0s - loss: 0.0517 - accuracy: 0.9930
Epoch 2: val_accuracy did not improve from 0.93575
21/21 [==============================] - 20s 970ms/step - loss: 0.0517 - accuracy: 0.9930 - val_loss: 0.2542 - val_accuracy: 0.9346
Epoch 3/50
21/21 [==============================] - ETA: 0s - loss: 0.0501 - accuracy: 0.9876
Epoch 3: val_accuracy did not improve from 0.93575
21/21 [==============================] - 21s 972ms/step - loss: 0.0501 - accuracy: 0.9876 - val_loss: 0.2614 - val_accuracy: 0.9346
Epoch 4/50
21/21 [==============================] - ETA: 0s - loss: 0.0444 - accuracy: 0.9946
Epoch 4: val_accuracy improved from 0.93575 to 0.94159, saving model to best_model.h5
21/21 [==============================] - 22s 1s/step - loss: 0.0444 - accuracy: 0.9946 - val_loss: 0.2528 - val_accuracy: 0.9416
Epoch 5/50
21/21 [==============================] - ETA: 0s - loss: 0.0408 - accuracy: 0.9930
Epoch 5: val_accuracy did not improve from 0.94159
21/21 [==============================] - 20s 972ms/step - loss: 0.0408 - accuracy: 0.9930 - val_loss: 0.2612 - val_accuracy: 0.9393
Epoch 6/50
21/21 [==============================] - ETA: 0s - loss: 0.0451 - accuracy: 0.9891
Epoch 6: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 982ms/step - loss: 0.0451 - accuracy: 0.9891 - val_loss: 0.2579 - val_accuracy: 0.9393
Epoch 7/50
21/21 [==============================] - ETA: 0s - loss: 0.0426 - accuracy: 0.9907
Epoch 7: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 980ms/step - loss: 0.0426 - accuracy: 0.9907 - val_loss: 0.2565 - val_accuracy: 0.9357
Epoch 8/50
21/21 [==============================] - ETA: 0s - loss: 0.0359 - accuracy: 0.9946
Epoch 8: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 1s/step - loss: 0.0359 - accuracy: 0.9946 - val_loss: 0.2613 - val_accuracy: 0.9357
Epoch 9/50
21/21 [==============================] - ETA: 0s - loss: 0.0341 - accuracy: 0.9946
Epoch 9: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 995ms/step - loss: 0.0341 - accuracy: 0.9946 - val_loss: 0.3145 - val_accuracy: 0.9089
Epoch 10/50
21/21 [==============================] - ETA: 0s - loss: 0.0973 - accuracy: 0.9627
Epoch 10: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 994ms/step - loss: 0.0973 - accuracy: 0.9627 - val_loss: 0.2960 - val_accuracy: 0.9171
Epoch 11/50
21/21 [==============================] - ETA: 0s - loss: 0.0514 - accuracy: 0.9852
Epoch 11: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 1s/step - loss: 0.0514 - accuracy: 0.9852 - val_loss: 0.2785 - val_accuracy: 0.9346
Epoch 12/50
21/21 [==============================] - ETA: 0s - loss: 0.0358 - accuracy: 0.9938
Epoch 12: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0358 - accuracy: 0.9938 - val_loss: 0.2683 - val_accuracy: 0.9334
Epoch 13/50
21/21 [==============================] - ETA: 0s - loss: 0.0299 - accuracy: 0.9953
Epoch 13: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0299 - accuracy: 0.9953 - val_loss: 0.2756 - val_accuracy: 0.9334
Epoch 14/50
21/21 [==============================] - ETA: 0s - loss: 0.0323 - accuracy: 0.9930
Epoch 14: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0323 - accuracy: 0.9930 - val_loss: 0.2870 - val_accuracy: 0.9334
Epoch 15/50
21/21 [==============================] - ETA: 0s - loss: 0.0274 - accuracy: 0.9969
Epoch 15: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0274 - accuracy: 0.9969 - val_loss: 0.2787 - val_accuracy: 0.9357
Epoch 16/50
21/21 [==============================] - ETA: 0s - loss: 0.0250 - accuracy: 0.9969
Epoch 16: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0250 - accuracy: 0.9969 - val_loss: 0.2853 - val_accuracy: 0.9346
Epoch 17/50
21/21 [==============================] - ETA: 0s - loss: 0.0267 - accuracy: 0.9969
Epoch 17: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0267 - accuracy: 0.9969 - val_loss: 0.2888 - val_accuracy: 0.9346
Epoch 18/50
21/21 [==============================] - ETA: 0s - loss: 0.0255 - accuracy: 0.9984
Epoch 18: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0255 - accuracy: 0.9984 - val_loss: 0.2883 - val_accuracy: 0.9346
Epoch 19/50
21/21 [==============================] - ETA: 0s - loss: 0.0222 - accuracy: 0.9977
Epoch 19: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0222 - accuracy: 0.9977 - val_loss: 0.2876 - val_accuracy: 0.9311
Epoch 20/50
21/21 [==============================] - ETA: 0s - loss: 0.0229 - accuracy: 0.9984
Epoch 20: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0229 - accuracy: 0.9984 - val_loss: 0.2878 - val_accuracy: 0.9322
Epoch 21/50
21/21 [==============================] - ETA: 0s - loss: 0.0214 - accuracy: 0.9961
Epoch 21: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0214 - accuracy: 0.9961 - val_loss: 0.2841 - val_accuracy: 0.9393
Epoch 22/50
21/21 [==============================] - ETA: 0s - loss: 0.0218 - accuracy: 0.9977
Epoch 22: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0218 - accuracy: 0.9977 - val_loss: 0.3001 - val_accuracy: 0.9311
Epoch 23/50
21/21 [==============================] - ETA: 0s - loss: 0.0244 - accuracy: 0.9961
Epoch 23: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0244 - accuracy: 0.9961 - val_loss: 0.3012 - val_accuracy: 0.9334
Epoch 24/50
21/21 [==============================] - ETA: 0s - loss: 0.0175 - accuracy: 0.9984
Epoch 24: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0175 - accuracy: 0.9984 - val_loss: 0.2951 - val_accuracy: 0.9322
Epoch 25/50
21/21 [==============================] - ETA: 0s - loss: 0.0160 - accuracy: 0.9992
Epoch 25: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0160 - accuracy: 0.9992 - val_loss: 0.3044 - val_accuracy: 0.9287
Epoch 26/50
21/21 [==============================] - ETA: 0s - loss: 0.0190 - accuracy: 0.9984
Epoch 26: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 1s/step - loss: 0.0190 - accuracy: 0.9984 - val_loss: 0.3013 - val_accuracy: 0.9357
Epoch 27/50
21/21 [==============================] - ETA: 0s - loss: 0.0171 - accuracy: 1.0000
Epoch 27: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 999ms/step - loss: 0.0171 - accuracy: 1.0000 - val_loss: 0.3000 - val_accuracy: 0.9369
Epoch 28/50
21/21 [==============================] - ETA: 0s - loss: 0.0179 - accuracy: 0.9984
Epoch 28: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0179 - accuracy: 0.9984 - val_loss: 0.3057 - val_accuracy: 0.9334
Epoch 29/50
21/21 [==============================] - ETA: 0s - loss: 0.0149 - accuracy: 1.0000
Epoch 29: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0149 - accuracy: 1.0000 - val_loss: 0.3206 - val_accuracy: 0.9252
Epoch 30/50
21/21 [==============================] - ETA: 0s - loss: 0.0240 - accuracy: 0.9930
Epoch 30: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0240 - accuracy: 0.9930 - val_loss: 0.3077 - val_accuracy: 0.9276
Epoch 31/50
21/21 [==============================] - ETA: 0s - loss: 0.0347 - accuracy: 0.9953
Epoch 31: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 998ms/step - loss: 0.0347 - accuracy: 0.9953 - val_loss: 0.3073 - val_accuracy: 0.9322
Epoch 32/50
21/21 [==============================] - ETA: 0s - loss: 0.0206 - accuracy: 0.9961
Epoch 32: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0206 - accuracy: 0.9961 - val_loss: 0.3000 - val_accuracy: 0.9334
Epoch 33/50
21/21 [==============================] - ETA: 0s - loss: 0.0211 - accuracy: 0.9984
Epoch 33: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0211 - accuracy: 0.9984 - val_loss: 0.3100 - val_accuracy: 0.9322
Epoch 34/50
21/21 [==============================] - ETA: 0s - loss: 0.0208 - accuracy: 0.9953
Epoch 34: val_accuracy did not improve from 0.94159
21/21 [==============================] - 22s 1s/step - loss: 0.0208 - accuracy: 0.9953 - val_loss: 0.3120 - val_accuracy: 0.9287
Epoch 35/50
21/21 [==============================] - ETA: 0s - loss: 0.0371 - accuracy: 0.9868
Epoch 35: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 987ms/step - loss: 0.0371 - accuracy: 0.9868 - val_loss: 0.3768 - val_accuracy: 0.9136
Epoch 36/50
21/21 [==============================] - ETA: 0s - loss: 0.0257 - accuracy: 0.9961
Epoch 36: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 1s/step - loss: 0.0257 - accuracy: 0.9961 - val_loss: 0.3138 - val_accuracy: 0.9276
Epoch 37/50
21/21 [==============================] - ETA: 0s - loss: 0.0126 - accuracy: 0.9992
Epoch 37: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 997ms/step - loss: 0.0126 - accuracy: 0.9992 - val_loss: 0.3177 - val_accuracy: 0.9322
Epoch 38/50
21/21 [==============================] - ETA: 0s - loss: 0.0116 - accuracy: 0.9984
Epoch 38: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 988ms/step - loss: 0.0116 - accuracy: 0.9984 - val_loss: 0.3296 - val_accuracy: 0.9299
Epoch 39/50
21/21 [==============================] - ETA: 0s - loss: 0.0105 - accuracy: 1.0000
Epoch 39: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 974ms/step - loss: 0.0105 - accuracy: 1.0000 - val_loss: 0.3263 - val_accuracy: 0.9311
Epoch 40/50
21/21 [==============================] - ETA: 0s - loss: 0.0109 - accuracy: 0.9992
Epoch 40: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 977ms/step - loss: 0.0109 - accuracy: 0.9992 - val_loss: 0.3390 - val_accuracy: 0.9276
Epoch 41/50
21/21 [==============================] - ETA: 0s - loss: 0.0158 - accuracy: 0.9977
Epoch 41: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 993ms/step - loss: 0.0158 - accuracy: 0.9977 - val_loss: 0.3146 - val_accuracy: 0.9287
Epoch 42/50
21/21 [==============================] - ETA: 0s - loss: 0.0133 - accuracy: 0.9984
Epoch 42: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 994ms/step - loss: 0.0133 - accuracy: 0.9984 - val_loss: 0.3648 - val_accuracy: 0.9182
Epoch 43/50
21/21 [==============================] - ETA: 0s - loss: 0.0120 - accuracy: 0.9992
Epoch 43: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 1s/step - loss: 0.0120 - accuracy: 0.9992 - val_loss: 0.3235 - val_accuracy: 0.9311
Epoch 44/50
21/21 [==============================] - ETA: 0s - loss: 0.0103 - accuracy: 0.9984
Epoch 44: val_accuracy did not improve from 0.94159
21/21 [==============================] - 21s 985ms/step - loss: 0.0103 - accuracy: 0.9984 - val_loss: 0.3347 - val_accuracy: 0.9299
Epoch 45/50
21/21 [==============================] - ETA: 0s - loss: 0.0097 - accuracy: 0.9992
Epoch 45: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0097 - accuracy: 0.9992 - val_loss: 0.3403 - val_accuracy: 0.9299
Epoch 46/50
21/21 [==============================] - ETA: 0s - loss: 0.0103 - accuracy: 0.9992
Epoch 46: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0103 - accuracy: 0.9992 - val_loss: 0.3351 - val_accuracy: 0.9264
Epoch 47/50
21/21 [==============================] - ETA: 0s - loss: 0.0078 - accuracy: 1.0000
Epoch 47: val_accuracy did not improve from 0.94159
21/21 [==============================] - 23s 1s/step - loss: 0.0078 - accuracy: 1.0000 - val_loss: 0.3426 - val_accuracy: 0.9311
Epoch 48/50
21/21 [==============================] - ETA: 0s - loss: 0.0095 - accuracy: 0.9992
Epoch 48: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0095 - accuracy: 0.9992 - val_loss: 0.3427 - val_accuracy: 0.9287
Epoch 49/50
21/21 [==============================] - ETA: 0s - loss: 0.0084 - accuracy: 0.9992
Epoch 49: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0084 - accuracy: 0.9992 - val_loss: 0.3540 - val_accuracy: 0.9311
Epoch 50/50
21/21 [==============================] - ETA: 0s - loss: 0.0083 - accuracy: 1.0000
Epoch 50: val_accuracy did not improve from 0.94159
21/21 [==============================] - 24s 1s/step - loss: 0.0083 - accuracy: 1.0000 - val_loss: 0.3445 - val_accuracy: 0.9311

五、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("./45-data/Monkeypox/M06_01_04.jpg")  #这里选择你需要预测的图片
img = Image.open("./ills/Others/NM15_02_11.jpg")  #这里选择你需要预测的图片
image = tf.image.resize(img, [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)])

1/1 [==============================] - 0s 27ms/step
预测结果为： Others

心得：

第一次训练的时候出现了过拟合的情况，故增大了batch_size的数量，适当的减少了训练集的数量。故训练集和测试集准确率相差再5%之内

小叮当爱咖啡

关注

16
点赞
踩
11

收藏

觉得还不错? 一键收藏
0
评论
Tensorflow实现深度学习案例4：猴痘病识别

本文为为内部文章原作者：K同学啊。
复制链接

扫一扫

Tensorflow实现深度学习案例4：猴痘病识别

一 前期准备

1.导入数据

2.查看数据

3.数据预处理

4.再次查看数据

5. 配置数据集

二、构建CNN网络

三、编译

四、训练模型

五、模型评估

1. Loss与Accuracy图

2. 指定图片进行预测

心得：

一前期准备