学习日记
1,学习知识点
卷积神经网络(CNN)天气识别
2,学习遇到的问题
内容较复杂,难懂
3,学习的收获
采用CNN实现多云、下雨、晴、日出四种天气状态的识别。本文为了增加模型的泛化能力,新增了Dropout层并且将最大池化层调整成了平均池化层。
4,实操
- 语言环境:Python3.6.5
- 编译器:jupyter notebook
- 深度学习环境:TensorFlow2
1. 设置GPU
如果使用的是CPU可以忽略这步
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")
2. 导入数据
import matplotlib.pyplot as plt import os,PIL # 设置随机种子尽可能使结果可以重现 import numpy as np np.random.seed(1) # 设置随机种子尽可能使结果可以重现 import tensorflow as tf tf.random.set_seed(1) from tensorflow import keras from tensorflow.keras import layers,models import pathlibdata_dir = "D:/jupyter notebook/DL-100-days/datasets/weather_photos/" data_dir = pathlib.Path(data_dir)
3. 查看数据
数据集一共分为
cloudy
、rain
、shine
、sunrise
四类,分别存放于weather_photos
文件夹中以各自名字命名的子文件夹中。image_count = len(list(data_dir.glob('*/*.jpg'))) print("图片总数为:",image_count)图片总数为: 1125roses = list(data_dir.glob('sunrise/*.jpg')) PIL.Image.open(str(roses[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batch_size = 32 img_height = 180 img_width = 180""" 关于image_dataset_from_directory()的详细介绍可以参考文章: """ 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 1125 files belonging to 4 classes. Using 900 files for training.""" 关于image_dataset_from_directory()的详细介绍可以参考文章: """ 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 1125 files belonging to 4 classes. Using 225 files for validation.
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)['cloudy', 'rain', 'shine', 'sunrise']
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, 180, 180, 3) (32,)
Image_batch
是形状的张量(32,180,180,3)。这是一批形状180x180x3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: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.AUTOTUNEtrain_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,fashion_mnist 数据集中的图片,形状是(28, 28, 1)
即灰度图像。我们需要在声明第一层时将形状赋值给参数input_shape
。num_classes = 4""" 关于卷积核的计算不懂的可以参考文章: layers.Dropout(0.4) 作用是防止过拟合,提高模型的泛化能力。 在上一篇文章花朵识别中,训练准确率与验证准确率相差巨大就是由于模型过拟合导致的 关于Dropout层的更多介绍可以参考文章: """ 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.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, 180, 180, 3) 0 _________________________________________________________________ conv2d (Conv2D) (None, 178, 178, 16) 448 _________________________________________________________________ average_pooling2d (AveragePo (None, 89, 89, 16) 0 _________________________________________________________________ conv2d_1 (Conv2D) (None, 87, 87, 32) 4640 _________________________________________________________________ average_pooling2d_1 (Average (None, 43, 43, 32) 0 _________________________________________________________________ conv2d_2 (Conv2D) (None, 41, 41, 64) 18496 _________________________________________________________________ dropout (Dropout) (None, 41, 41, 64) 0 _________________________________________________________________ flatten (Flatten) (None, 107584) 0 _________________________________________________________________ dense (Dense) (None, 128) 13770880 _________________________________________________________________ dense_1 (Dense) (None, 5) 645 ================================================================= Total params: 13,795,109 Trainable params: 13,795,109 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。
# 设置优化器opt = tf.keras.optimizers.Adam(learning_rate=0.001) model.compile(optimizer=opt, loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True), metrics=['accuracy'])
五、训练模型
epochs = 10history = model.fit( train_ds, validation_data=val_ds, epochs=epochs )Epoch 1/1029/29 [==============================] - 6s 58ms/step - loss: 1.5865 - accuracy: 0.4463 - val_loss: 0.5837 - val_accuracy: 0.7689 Epoch 2/10 29/29 [==============================] - 0s 12ms/step - loss: 0.5289 - accuracy: 0.8295 - val_loss: 0.5405 - val_accuracy: 0.8133 Epoch 3/10 29/29 [==============================] - 0s 12ms/step - loss: 0.2930 - accuracy: 0.8967 - val_loss: 0.5364 - val_accuracy: 0.8000 Epoch 4/10 29/29 [==============================] - 0s 12ms/step - loss: 0.2742 - accuracy: 0.9074 - val_loss: 0.4034 - val_accuracy: 0.8267 Epoch 5/10 29/29 [==============================] - 0s 11ms/step - loss: 0.1952 - accuracy: 0.9383 - val_loss: 0.3874 - val_accuracy: 0.8844 Epoch 6/10 29/29 [==============================] - 0s 11ms/step - loss: 0.1592 - accuracy: 0.9468 - val_loss: 0.3680 - val_accuracy: 0.8756 Epoch 7/10 29/29 [==============================] - 0s 12ms/step - loss: 0.0836 - accuracy: 0.9755 - val_loss: 0.3429 - val_accuracy: 0.8756 Epoch 8/10 29/29 [==============================] - 0s 12ms/step - loss: 0.0943 - accuracy: 0.9692 - val_loss: 0.3836 - val_accuracy: 0.9067 Epoch 9/10 29/29 [==============================] - 0s 12ms/step - loss: 0.0344 - accuracy: 0.9909 - val_loss: 0.3578 - val_accuracy: 0.9067 Epoch 10/10 29/29 [==============================] - 0s 11ms/step - loss: 0.0950 - accuracy: 0.9708 - val_loss: 0.4710 - val_accuracy: 0.8356
六、模型评估
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()