DAY 43 复习日

import tensorflow as tf
from tensorflow.keras import datasets, layers, models
import matplotlib.pyplot as plt# 加载CIFAR10数据集
(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()# 将像素值归一化到0和1之间
train_images, test_images = train_images / 255.0, test_images / 255.0# 类别名称
class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer','dog', 'frog', 'horse', 'ship', 'truck']

model = models.Sequential()
# 第一个卷积-池化层
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))
model.add(layers.MaxPooling2D((2, 2)))
# 第二个卷积-池化层
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
# 第三个卷积层
model.add(layers.Conv2D(64, (3, 3), activation='relu'))# 添加全连接层进行分类
model.add(layers.Flatten())
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(10)) # 输出层有10个节点，对应10个类别model.summary()

# 编译模型
model.compile(optimizer='adam',loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),metrics=['accuracy'])# 训练模型
history = model.fit(train_images, train_labels, epochs=10, validation_data=(test_images, test_labels))

import numpy as np
import tensorflow as tf
import cv2def make_gradcam_heatmap(img_array, model, last_conv_layer_name, pred_index=None):# 创建一个子模型，输出最后一个卷积层的激活图和模型的最终预测grad_model = tf.keras.models.Model([model.inputs], [model.get_layer(last_conv_layer_name).output, model.output])# 使用GradientTape记录梯度计算过程with tf.GradientTape() as tape:last_conv_layer_output, preds = grad_model(img_array)if pred_index is None:pred_index = tf.argmax(preds[0])class_channel = preds[:, pred_index]# 计算预测类别对于最后一个卷积层输出的梯度grads = tape.gradient(class_channel, last_conv_layer_output)# 计算每个通道的平均梯度，作为权重pooled_grads = tf.reduce_mean(grads, axis=(0, 1, 2))# 将权重与特征图相乘，得到热力图last_conv_layer_output = last_conv_layer_output[0]heatmap = last_conv_layer_output @ pooled_grads[..., tf.newaxis]heatmap = tf.squeeze(heatmap)# 归一化热力图heatmap = tf.maximum(heatmap, 0) / tf.math.reduce_max(heatmap)return heatmap.numpy()def display_gradcam(img, heatmap, alpha=0.4):# 将热力图调整为图像大小heatmap = cv2.resize(heatmap, (img.shape[1], img.shape[0]))heatmap = np.uint8(255 * heatmap)heatmap = cv2.applyColorMap(heatmap, cv2.COLORMAP_JET)# 将热力图叠加到原始图像上superimposed_img = heatmap * alpha + img * (1-alpha)superimposed_img = np.clip(superimposed_img, 0, 255).astype(np.uint8)return superimposed_img