第九周 tensorflow实现猫狗识别2

最新推荐文章于 2023-11-05 13:31:52 发布
最新推荐文章于 2023-11-05 13:31:52 发布
阅读量802 收藏 2
点赞数
分类专栏： 365天深度学习 tensorflow系列 文章标签： tensorflow 人工智能 python
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/qq_44753643/article/details/129978000
版权
本文介绍了使用TensorFlow进行猫狗图像分类的实践过程，包括数据预处理、数据增强、构建卷积神经网络（CNN）、模型训练以及使用tqdm实现进度条可视化。通过8周的学习，优化了模型，提高了识别准确率。
摘要由CSDN通过智能技术生成

一、本周学习内容:

1、找到第八周的问题

第8周的问题处在训练集和测试集数据的历史损失和准确率，因为是每一个batch输出一次history而不是epoch输出一次，因此我们在最终输出时对一个epoch中的batch取平均值

2、数据增强

from tensorflow.keras.preprocessing.image import ImageDataGenerator
ImageDataGenerator（
rescale=所有数据集将乘以该数值,
rotation_range=随即旋转角度数范围,
width_shift_range=随即宽度偏移量,
height_shift_range=随即高度偏移量,
horizontal_flip=是否随机水平翻转,
zoom_range=随机缩放的范围 -> [1-n,1+n]
）
参考这个大佬的链接

3、了解tqdm，并使用tqdm实现可视化进度条

tqdm是 Python 进度条库，可以在 Python长循环中添加一个进度提示信息。用户只需要封装任意的迭代器，是一个快速、扩展性强的进度条工具库。
这篇文章介绍tqdm的使用

二、前言

我们的数据图片共有3400张,两个类别,类别即为文件夹名。
类别包括:[‘cat’, ‘dog’]

三、电脑环境

电脑系统：Windows 10
语言环境：Python 3.8.8
编译器：Pycharm 2021.1.3
深度学习环境：TensorFlow 2.8.0,keras 2.8.0
显卡及显存：RTX 3070 8G

四、前期准备

1、导入相关依赖项

from keras.models import Sequential
from keras.layers import *
from tensorflow import keras
from keras.callbacks import ModelCheckpoint
import tensorflow as tf
import matplotlib.pyplot as plt

2、设置GPU(我下载的tensorflow-gpu 默认使用GPU)

只使用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")

使用cpu和gpu
os.environ[“CUDA_VISIBLE_DEVICES”] = “-1”

3、加载数据集和展示

(1)、数据预处理

# 创建数据增强器
train_datagen=ImageDataGenerator(
    rescale=1. / 255,
    rotation_range=45,
    width_shift_range=.15,
    height_shift_range=.15,
    horizontal_flip=False,
    zoom_range=0.5
)
val_datagen=ImageDataGenerator(
    rescale=1. / 255,
    rotation_range=45,
    width_shift_range=.15,
    height_shift_range=.15,
    horizontal_flip=False,
    zoom_range=0.5
)
# 加载数据
train_ds=train_datagen.flow_from_directory(
    data_dir,
    target_size=(224,224),
    batch_size=10,  #迭代一次数据大小
    class_mode='categorical'#如果One-hot就是categorical,二分类用binary
)

val_ds=val_datagen.flow_from_directory(
    data_dir,
    target_size=(224,224),
    batch_size=20,
    class_mode='categorical'
)
# 获取标签
class_names = list(train_generator.class_indices.keys())

print(class_names)

(2)、数据展示

图片展示

# 图片展示
for images, labels in train_ds:
    for i in range(10):
        ax = plt.subplot(5, 10, i + 1)
        plt.imshow(images[i])
        print(labels)
        print(len(images))
        plt.title(class_names[np.argmax(labels[i])])
        plt.axis("off")
        break
    break

plt.show()  #使用pycharm的需要加入这行代码才能将图像显示出来

在这里插入图片描述

五、搭建CNN网络

def vgg_net(num_layer,filter):
    blk = Sequential()
    for _ in range(num_layer):
        blk.add(Conv2D(filters=filter,kernel_size=3,padding='same',activation='relu'))
    blk.add(MaxPool2D(pool_size=(2,2),strides=2))
    return blk

conv_arch = [(2,64),(2,128),(3,256),(3,512),(3,512)]
# model = Sequential([Conv2D(filters=64,kernel_size=3,padding='same',activation='relu',input_shape=(224,224,1))])
model = Sequential()
for num_layer,filter in conv_arch:
    model.add(vgg_net(num_layer,filter))
# print(model.summary())
model.add(Sequential([
    Flatten(),
    Dense(4096,activation='relu'),
    Dropout(0.5),
    Dense(4096,activation='relu'),
    Dropout(0.5),
    Dense(len(class_names),activation='softmax')
]))

x = tf.random.uniform((1,224,224,3))
y = model(x)


model.compile(optimizer="adam",
              loss     ='categorical_crossentropy',
              metrics  =['accuracy'])


from tqdm import tqdm
import tensorflow.keras.backend as K

epochs = 10
lr = 1e-4

# 记录训练数据，方便后面的分析
history_train_loss = []
history_train_accuracy = []
history_val_loss = []
history_val_accuracy = []

for epoch in range(epochs):
    train_total = len(train_ds)
    val_total = len(val_ds)

    """
    total：预期的迭代数目
    ncols：控制进度条宽度
    mininterval：进度更新最小间隔，以秒为单位（默认值：0.1）
    """
    with tqdm(total=train_total, desc=f'Epoch {epoch + 1}/{epochs}', mininterval=1, ncols=100) as pbar:

        lr = lr * 0.92
        K.set_value(model.optimizer.lr, lr)
        train_loss = []
        train_accuracy = []
        for image, label in train_ds:
            """
            训练模型，简单理解train_on_batch就是：它是比model.fit()更高级的一个用法
            想详细了解 train_on_batch 的同学，
            可以看看这篇文章：https://www.yuque.com/mingtian-fkmxf/hv4lcq/ztt4gy
            """
            history = model.train_on_batch(image, label)

            train_loss.append(history[0])
            train_accuracy.append(history[1])

            pbar.set_postfix({"loss": "%.4f" % history[0],
                              "accuracy": "%.4f" % history[1],
                              "lr": K.get_value(model.optimizer.lr)})
            pbar.update(1)
        history_train_loss.append(np.mean(train_loss))
        history_train_accuracy.append(np.mean(train_accuracy))
        print('开始验证！')

        with tqdm(total=val_total, desc=f'Epoch {epoch + 1}/{epochs}', mininterval=0.3, ncols=100) as pbar:
            val_loss=[]
            val_accuracy = []
            for image, label in val_ds:
                history = model.test_on_batch(image, label)

                val_loss.append(history[0])
                val_accuracy.append(history[1])

                pbar.set_postfix({"loss": "%.4f" % history[0],
                                  "accuracy": "%.4f" % history[1]})
                pbar.update(1)
            history_val_loss.append(np.mean(val_loss))
            history_val_accuracy.append(np.mean(val_accuracy))

        print('结束验证！')
        print("验证loss为：%.4f" % np.mean(val_loss))
        print("验证准确率为：%.4f" % np.mean(val_accuracy))

Epoch 1/10: 100%|██████████| 55/55 [00:55<00:00,  1.25it/s, loss=0.7950, accuracy=0.5200, lr=9.2e-5]
开始验证！
Epoch 1/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]
Epoch 1/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.6692, accuracy=0.5400]
Epoch 1/10:   7%|█▌                    | 1/14 [00:00<00:07,  1.67it/s, loss=0.6692, accuracy=0.5400]
Epoch 1/10:   7%|█▌                    | 1/14 [00:00<00:07,  1.67it/s, loss=0.6247, accuracy=0.6600]
Epoch 1/10:  14%|███▏                  | 2/14 [00:00<00:07,  1.67it/s, loss=0.6243, accuracy=0.6200]
Epoch 1/10:  21%|████▋                 | 3/14 [00:01<00:06,  1.67it/s, loss=0.5732, accuracy=0.7600]
Epoch 1/10:  29%|██████▎               | 4/14 [00:01<00:02,  4.38it/s, loss=0.5732, accuracy=0.7600]
Epoch 1/10:  29%|██████▎               | 4/14 [00:01<00:02,  4.38it/s, loss=0.5970, accuracy=0.6600]
Epoch 1/10:  36%|███████▊              | 5/14 [00:01<00:02,  4.38it/s, loss=0.5614, accuracy=0.7400]
Epoch 1/10:  43%|█████████▍            | 6/14 [00:01<00:01,  4.38it/s, loss=0.5716, accuracy=0.6800]
Epoch 1/10:  50%|███████████           | 7/14 [00:01<00:01,  5.49it/s, loss=0.5716, accuracy=0.6800]
Epoch 1/10:  50%|███████████           | 7/14 [00:01<00:01,  5.49it/s, loss=0.5993, accuracy=0.7200]
Epoch 1/10:  57%|████████████▌         | 8/14 [00:01<00:01,  5.49it/s, loss=0.5803, accuracy=0.7400]
Epoch 1/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  5.49it/s, loss=0.4826, accuracy=0.8600]
Epoch 1/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.02it/s, loss=0.4826, accuracy=0.8600]
Epoch 1/10:  71%|███████████████      | 10/14 [00:02<00:00,  6.02it/s, loss=0.6271, accuracy=0.5800]
Epoch 1/10:  79%|████████████████▌    | 11/14 [00:02<00:00,  6.02it/s, loss=0.5784, accuracy=0.6600]
Epoch 1/10:  86%|██████████████████   | 12/14 [00:02<00:00,  6.02it/s, loss=0.6444, accuracy=0.6800]
Epoch 1/10:  93%|███████████████████▌ | 13/14 [00:02<00:00,  6.34it/s, loss=0.6444, accuracy=0.6800]
Epoch 1/10: 100%|█████████████████████| 14/14 [00:05<00:00,  2.55it/s, loss=0.6598, accuracy=0.6000]
Epoch 1/10: 100%|██████████| 55/55 [01:01<00:00,  1.11s/it, loss=0.7950, accuracy=0.5200, lr=9.2e-5]
Epoch 2/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.5995
验证准确率为：0.6786
Epoch 2/10:  98%|████████▊| 54/55 [00:26<00:00,  2.05it/s, loss=0.3791, accuracy=0.9000, lr=8.46e-5]
开始验证！
Epoch 2/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]
Epoch 2/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.2396, accuracy=0.9200]
Epoch 2/10:   7%|█▌                    | 1/14 [00:00<00:03,  3.43it/s, loss=0.2423, accuracy=0.9200]
Epoch 2/10:  14%|███▏                  | 2/14 [00:00<00:02,  4.57it/s, loss=0.2737, accuracy=0.8800]
Epoch 2/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.85it/s, loss=0.2737, accuracy=0.8800]
Epoch 2/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.85it/s, loss=0.1539, accuracy=0.9000]
Epoch 2/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.85it/s, loss=0.2980, accuracy=0.9200]
Epoch 2/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.85it/s, loss=0.1904, accuracy=0.9200]
Epoch 2/10:  43%|█████████▍            | 6/14 [00:00<00:01,  6.86it/s, loss=0.1904, accuracy=0.9200]
Epoch 2/10:  43%|█████████▍            | 6/14 [00:01<00:01,  6.86it/s, loss=0.1032, accuracy=0.9600]
Epoch 2/10:  50%|███████████           | 7/14 [00:01<00:01,  6.86it/s, loss=0.1594, accuracy=0.9400]
Epoch 2/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.86it/s, loss=0.2330, accuracy=0.9600]
Epoch 2/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.90it/s, loss=0.2330, accuracy=0.9600]
Epoch 2/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.90it/s, loss=0.1394, accuracy=0.9800]
Epoch 2/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.90it/s, loss=0.2391, accuracy=0.8800]
Epoch 2/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.90it/s, loss=0.2008, accuracy=0.9000]
Epoch 2/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.94it/s, loss=0.2008, accuracy=0.9000]
Epoch 2/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.94it/s, loss=0.3320, accuracy=0.8600]
Epoch 2/10: 100%|█████████████████████| 14/14 [00:01<00:00,  7.12it/s, loss=0.1901, accuracy=0.9000]
Epoch 2/10: 100%|█████████| 55/55 [00:28<00:00,  1.94it/s, loss=0.3791, accuracy=0.9000, lr=8.46e-5]
Epoch 3/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.2139
验证准确率为：0.9171
Epoch 3/10:  98%|████████▊| 54/55 [00:26<00:00,  2.04it/s, loss=0.0466, accuracy=0.9800, lr=7.79e-5]
Epoch 3/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]开始验证！

Epoch 3/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.1344, accuracy=0.9000]
Epoch 3/10:   7%|█▌                    | 1/14 [00:00<00:03,  3.42it/s, loss=0.1919, accuracy=0.9200]
Epoch 3/10:  14%|███▏                  | 2/14 [00:00<00:02,  4.59it/s, loss=0.2520, accuracy=0.9200]
Epoch 3/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.87it/s, loss=0.2520, accuracy=0.9200]
Epoch 3/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.87it/s, loss=0.0580, accuracy=1.0000]
Epoch 3/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.87it/s, loss=0.0744, accuracy=0.9600]
Epoch 3/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.87it/s, loss=0.1914, accuracy=0.9200]
Epoch 3/10:  43%|█████████▍            | 6/14 [00:00<00:01,  6.81it/s, loss=0.1914, accuracy=0.9200]
Epoch 3/10:  43%|█████████▍            | 6/14 [00:01<00:01,  6.81it/s, loss=0.0322, accuracy=1.0000]
Epoch 3/10:  50%|███████████           | 7/14 [00:01<00:01,  6.81it/s, loss=0.0940, accuracy=0.9800]
Epoch 3/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.81it/s, loss=0.1493, accuracy=0.9400]
Epoch 3/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.87it/s, loss=0.1493, accuracy=0.9400]
Epoch 3/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.87it/s, loss=0.0906, accuracy=0.9600]
Epoch 3/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.87it/s, loss=0.2302, accuracy=0.9000]
Epoch 3/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.87it/s, loss=0.1218, accuracy=0.9600]
Epoch 3/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.90it/s, loss=0.1218, accuracy=0.9600]
Epoch 3/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.90it/s, loss=0.2033, accuracy=0.9200]
Epoch 3/10: 100%|█████████████████████| 14/14 [00:01<00:00,  7.08it/s, loss=0.0882, accuracy=0.9333]
Epoch 3/10: 100%|█████████| 55/55 [00:28<00:00,  1.92it/s, loss=0.0466, accuracy=0.9800, lr=7.79e-5]
Epoch 4/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.1365
验证准确率为：0.9438
Epoch 4/10: 100%|█████████| 55/55 [00:27<00:00,  1.95it/s, loss=0.1313, accuracy=0.9600, lr=7.16e-5]
Epoch 4/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]开始验证！

Epoch 4/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.1331, accuracy=0.9400]
Epoch 4/10:   7%|█▌                    | 1/14 [00:00<00:04,  3.05it/s, loss=0.0935, accuracy=0.9800]
Epoch 4/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.11it/s, loss=0.0935, accuracy=0.9800]
Epoch 4/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.11it/s, loss=0.2275, accuracy=0.9400]
Epoch 4/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.11it/s, loss=0.0317, accuracy=1.0000]
Epoch 4/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.12it/s, loss=0.0317, accuracy=1.0000]
Epoch 4/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.12it/s, loss=0.1114, accuracy=0.9600]
Epoch 4/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.12it/s, loss=0.0937, accuracy=0.9400]
Epoch 4/10:  43%|█████████▍            | 6/14 [00:00<00:01,  6.26it/s, loss=0.0937, accuracy=0.9400]
Epoch 4/10:  43%|█████████▍            | 6/14 [00:01<00:01,  6.26it/s, loss=0.0238, accuracy=0.9800]
Epoch 4/10:  50%|███████████           | 7/14 [00:01<00:01,  6.26it/s, loss=0.0896, accuracy=0.9800]
Epoch 4/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.41it/s, loss=0.0896, accuracy=0.9800]
Epoch 4/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.41it/s, loss=0.1320, accuracy=0.9600]
Epoch 4/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.41it/s, loss=0.0116, accuracy=1.0000]
Epoch 4/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.41it/s, loss=0.1019, accuracy=0.9200]
Epoch 4/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.57it/s, loss=0.1019, accuracy=0.9200]
Epoch 4/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.57it/s, loss=0.1263, accuracy=0.9400]
Epoch 4/10:  86%|██████████████████   | 12/14 [00:02<00:00,  6.57it/s, loss=0.0810, accuracy=0.9200]
Epoch 4/10:  93%|███████████████████▌ | 13/14 [00:02<00:00,  6.57it/s, loss=0.0535, accuracy=0.9667]
Epoch 4/10: 100%|█████████████████████| 14/14 [00:02<00:00,  6.70it/s, loss=0.0535, accuracy=0.9667]
Epoch 4/10: 100%|█████████| 55/55 [00:29<00:00,  1.87it/s, loss=0.1313, accuracy=0.9600, lr=7.16e-5]
Epoch 5/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.0936
验证准确率为：0.9590
Epoch 5/10:  98%|████████▊| 54/55 [00:27<00:00,  1.94it/s, loss=0.0284, accuracy=0.9800, lr=6.59e-5]
Epoch 5/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]开始验证！

Epoch 5/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.1285, accuracy=0.9400]
Epoch 5/10:   7%|█▌                    | 1/14 [00:00<00:04,  2.92it/s, loss=0.0995, accuracy=0.9800]
Epoch 5/10:  14%|███▏                  | 2/14 [00:00<00:02,  5.85it/s, loss=0.0995, accuracy=0.9800]
Epoch 5/10:  14%|███▏                  | 2/14 [00:00<00:02,  5.85it/s, loss=0.1170, accuracy=0.9800]
Epoch 5/10:  21%|████▋                 | 3/14 [00:00<00:01,  5.85it/s, loss=0.0193, accuracy=1.0000]
Epoch 5/10:  29%|██████▎               | 4/14 [00:00<00:01,  5.85it/s, loss=0.0193, accuracy=1.0000]
Epoch 5/10:  29%|██████▎               | 4/14 [00:00<00:01,  5.85it/s, loss=0.0619, accuracy=0.9800]
Epoch 5/10:  36%|███████▊              | 5/14 [00:01<00:01,  5.85it/s, loss=0.0564, accuracy=0.9800]
Epoch 5/10:  43%|█████████▍            | 6/14 [00:01<00:01,  5.87it/s, loss=0.0564, accuracy=0.9800]
Epoch 5/10:  43%|█████████▍            | 6/14 [00:01<00:01,  5.87it/s, loss=0.0180, accuracy=1.0000]
Epoch 5/10:  50%|███████████           | 7/14 [00:01<00:01,  5.87it/s, loss=0.0712, accuracy=0.9600]
Epoch 5/10:  57%|████████████▌         | 8/14 [00:01<00:01,  5.90it/s, loss=0.0712, accuracy=0.9600]
Epoch 5/10:  57%|████████████▌         | 8/14 [00:01<00:01,  5.90it/s, loss=0.1603, accuracy=0.9200]
Epoch 5/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  5.90it/s, loss=0.0212, accuracy=1.0000]
Epoch 5/10:  71%|███████████████      | 10/14 [00:01<00:00,  5.94it/s, loss=0.0212, accuracy=1.0000]
Epoch 5/10:  71%|███████████████      | 10/14 [00:01<00:00,  5.94it/s, loss=0.1483, accuracy=0.9600]
Epoch 5/10:  79%|████████████████▌    | 11/14 [00:02<00:00,  5.94it/s, loss=0.0767, accuracy=0.9800]
Epoch 5/10:  86%|██████████████████   | 12/14 [00:02<00:00,  6.02it/s, loss=0.0767, accuracy=0.9800]
Epoch 5/10:  86%|██████████████████   | 12/14 [00:02<00:00,  6.02it/s, loss=0.0880, accuracy=0.9600]
Epoch 5/10: 100%|█████████████████████| 14/14 [00:02<00:00,  6.17it/s, loss=0.0126, accuracy=1.0000]
Epoch 5/10: 100%|█████████| 55/55 [00:30<00:00,  1.83it/s, loss=0.0284, accuracy=0.9800, lr=6.59e-5]
Epoch 6/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.0771
验证准确率为：0.9743
Epoch 6/10:  98%|████████▊| 54/55 [00:28<00:00,  1.92it/s, loss=0.0023, accuracy=1.0000, lr=6.06e-5]
Epoch 6/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]开始验证！

Epoch 6/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.1735, accuracy=0.9400]
Epoch 6/10:   7%|█▌                    | 1/14 [00:00<00:03,  3.28it/s, loss=0.0859, accuracy=0.9600]
Epoch 6/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.56it/s, loss=0.0859, accuracy=0.9600]
Epoch 6/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.56it/s, loss=0.1493, accuracy=0.9800]
Epoch 6/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.56it/s, loss=0.0028, accuracy=1.0000]
Epoch 6/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.56it/s, loss=0.0705, accuracy=0.9800]
Epoch 6/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.69it/s, loss=0.0705, accuracy=0.9800]
Epoch 6/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.69it/s, loss=0.0208, accuracy=1.0000]
Epoch 6/10:  43%|█████████▍            | 6/14 [00:01<00:01,  6.69it/s, loss=0.0109, accuracy=1.0000]
Epoch 6/10:  50%|███████████           | 7/14 [00:01<00:01,  6.69it/s, loss=0.0476, accuracy=0.9800]
Epoch 6/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.74it/s, loss=0.0476, accuracy=0.9800]
Epoch 6/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.74it/s, loss=0.2229, accuracy=0.9200]
Epoch 6/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.74it/s, loss=0.0016, accuracy=1.0000]
Epoch 6/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.74it/s, loss=0.0714, accuracy=0.9400]
Epoch 6/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.74it/s, loss=0.0714, accuracy=0.9400]
Epoch 6/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.74it/s, loss=0.0613, accuracy=0.9400]
Epoch 6/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.74it/s, loss=0.0545, accuracy=0.9800]
Epoch 6/10:  93%|███████████████████▌ | 13/14 [00:02<00:00,  6.74it/s, loss=0.0455, accuracy=0.9667]
Epoch 6/10: 100%|█████████████████████| 14/14 [00:02<00:00,  6.89it/s, loss=0.0455, accuracy=0.9667]
Epoch 6/10: 100%|█████████| 55/55 [00:30<00:00,  1.80it/s, loss=0.0023, accuracy=1.0000, lr=6.06e-5]
Epoch 7/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.0727
验证准确率为：0.9705
Epoch 7/10:  98%|████████▊| 54/55 [00:28<00:00,  2.06it/s, loss=0.0925, accuracy=0.9800, lr=5.58e-5]
Epoch 7/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]开始验证！

Epoch 7/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.1217, accuracy=0.9600]
Epoch 7/10:   7%|█▌                    | 1/14 [00:00<00:03,  3.28it/s, loss=0.0499, accuracy=0.9800]
Epoch 7/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.54it/s, loss=0.0499, accuracy=0.9800]
Epoch 7/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.54it/s, loss=0.1077, accuracy=0.9800]
Epoch 7/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.54it/s, loss=0.0084, accuracy=1.0000]
Epoch 7/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.60it/s, loss=0.0084, accuracy=1.0000]
Epoch 7/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.60it/s, loss=0.0317, accuracy=0.9800]
Epoch 7/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.60it/s, loss=0.0497, accuracy=0.9800]
Epoch 7/10:  43%|█████████▍            | 6/14 [00:00<00:01,  6.62it/s, loss=0.0497, accuracy=0.9800]
Epoch 7/10:  43%|█████████▍            | 6/14 [00:01<00:01,  6.62it/s, loss=0.0032, accuracy=1.0000]
Epoch 7/10:  50%|███████████           | 7/14 [00:01<00:01,  6.62it/s, loss=0.0116, accuracy=1.0000]
Epoch 7/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.62it/s, loss=0.1292, accuracy=0.9400]
Epoch 7/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.68it/s, loss=0.1292, accuracy=0.9400]
Epoch 7/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.68it/s, loss=0.0031, accuracy=1.0000]
Epoch 7/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.68it/s, loss=0.0670, accuracy=0.9800]
Epoch 7/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.68it/s, loss=0.0416, accuracy=0.9800]
Epoch 7/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.75it/s, loss=0.0416, accuracy=0.9800]
Epoch 7/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.75it/s, loss=0.0323, accuracy=0.9800]
Epoch 7/10: 100%|█████████████████████| 14/14 [00:02<00:00,  6.91it/s, loss=0.0175, accuracy=1.0000]
Epoch 7/10: 100%|█████████| 55/55 [00:30<00:00,  1.78it/s, loss=0.0925, accuracy=0.9800, lr=5.58e-5]
Epoch 8/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.0482
验证准确率为：0.9829
Epoch 8/10:  98%|████████▊| 54/55 [00:28<00:00,  1.91it/s, loss=0.0016, accuracy=1.0000, lr=5.13e-5]
Epoch 8/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]开始验证！

Epoch 8/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.1328, accuracy=0.9600]
Epoch 8/10:   7%|█▌                    | 1/14 [00:00<00:03,  3.32it/s, loss=0.0430, accuracy=0.9600]
Epoch 8/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.62it/s, loss=0.0430, accuracy=0.9600]
Epoch 8/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.62it/s, loss=0.1390, accuracy=0.9800]
Epoch 8/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.62it/s, loss=0.0029, accuracy=1.0000]
Epoch 8/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.64it/s, loss=0.0029, accuracy=1.0000]
Epoch 8/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.64it/s, loss=0.0137, accuracy=1.0000]
Epoch 8/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.64it/s, loss=0.0734, accuracy=0.9600]
Epoch 8/10:  43%|█████████▍            | 6/14 [00:01<00:01,  6.64it/s, loss=0.0021, accuracy=1.0000]
Epoch 8/10:  50%|███████████           | 7/14 [00:01<00:01,  6.65it/s, loss=0.0021, accuracy=1.0000]
Epoch 8/10:  50%|███████████           | 7/14 [00:01<00:01,  6.65it/s, loss=0.0032, accuracy=1.0000]
Epoch 8/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.65it/s, loss=0.2002, accuracy=0.9400]
Epoch 8/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.65it/s, loss=0.0039, accuracy=1.0000]
Epoch 8/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.65it/s, loss=0.0039, accuracy=1.0000]
Epoch 8/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.65it/s, loss=0.1951, accuracy=0.9400]
Epoch 8/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.65it/s, loss=0.1133, accuracy=0.9600]
Epoch 8/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.65it/s, loss=0.0187, accuracy=1.0000]
Epoch 8/10:  93%|███████████████████▌ | 13/14 [00:01<00:00,  6.69it/s, loss=0.0187, accuracy=1.0000]
Epoch 8/10: 100%|█████████████████████| 14/14 [00:02<00:00,  6.87it/s, loss=0.0072, accuracy=1.0000]
Epoch 8/10: 100%|█████████| 55/55 [00:30<00:00,  1.78it/s, loss=0.0016, accuracy=1.0000, lr=5.13e-5]
Epoch 9/10:   0%|                                                            | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.0677
验证准确率为：0.9786
Epoch 9/10: 100%|█████████| 55/55 [00:28<00:00,  1.88it/s, loss=0.1454, accuracy=0.9800, lr=4.72e-5]
开始验证！
Epoch 9/10:   0%|                                                            | 0/14 [00:00<?, ?it/s]
Epoch 9/10:   0%|                              | 0/14 [00:00<?, ?it/s, loss=0.0488, accuracy=0.9800]
Epoch 9/10:   7%|█▌                    | 1/14 [00:00<00:04,  3.21it/s, loss=0.0275, accuracy=0.9800]
Epoch 9/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.42it/s, loss=0.0275, accuracy=0.9800]
Epoch 9/10:  14%|███▏                  | 2/14 [00:00<00:01,  6.42it/s, loss=0.1349, accuracy=0.9800]
Epoch 9/10:  21%|████▋                 | 3/14 [00:00<00:01,  6.42it/s, loss=0.0158, accuracy=1.0000]
Epoch 9/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.45it/s, loss=0.0158, accuracy=1.0000]
Epoch 9/10:  29%|██████▎               | 4/14 [00:00<00:01,  6.45it/s, loss=0.0304, accuracy=0.9800]
Epoch 9/10:  36%|███████▊              | 5/14 [00:00<00:01,  6.45it/s, loss=0.0252, accuracy=1.0000]
Epoch 9/10:  43%|█████████▍            | 6/14 [00:00<00:01,  6.48it/s, loss=0.0252, accuracy=1.0000]
Epoch 9/10:  43%|█████████▍            | 6/14 [00:01<00:01,  6.48it/s, loss=0.0157, accuracy=0.9800]
Epoch 9/10:  50%|███████████           | 7/14 [00:01<00:01,  6.48it/s, loss=0.0130, accuracy=1.0000]
Epoch 9/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.50it/s, loss=0.0130, accuracy=1.0000]
Epoch 9/10:  57%|████████████▌         | 8/14 [00:01<00:00,  6.50it/s, loss=0.1652, accuracy=0.9400]
Epoch 9/10:  64%|██████████████▏       | 9/14 [00:01<00:00,  6.50it/s, loss=0.0035, accuracy=1.0000]
Epoch 9/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.53it/s, loss=0.0035, accuracy=1.0000]
Epoch 9/10:  71%|███████████████      | 10/14 [00:01<00:00,  6.53it/s, loss=0.0517, accuracy=0.9800]
Epoch 9/10:  79%|████████████████▌    | 11/14 [00:01<00:00,  6.53it/s, loss=0.0636, accuracy=0.9600]
Epoch 9/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.55it/s, loss=0.0636, accuracy=0.9600]
Epoch 9/10:  86%|██████████████████   | 12/14 [00:01<00:00,  6.55it/s, loss=0.0324, accuracy=0.9800]
Epoch 9/10: 100%|█████████████████████| 14/14 [00:02<00:00,  6.72it/s, loss=0.0075, accuracy=1.0000]
Epoch 9/10: 100%|█████████| 55/55 [00:30<00:00,  1.81it/s, loss=0.1454, accuracy=0.9800, lr=4.72e-5]
Epoch 10/10:   0%|                                                           | 0/55 [00:00<?, ?it/s]结束验证！
验证loss为：0.0454
验证准确率为：0.9829
Epoch 10/10: 100%|████████| 55/55 [00:28<00:00,  1.92it/s, loss=0.0038, accuracy=1.0000, lr=4.34e-5]
开始验证！
Epoch 10/10:   0%|                                                           | 0/14 [00:00<?, ?it/s]
Epoch 10/10:   0%|                             | 0/14 [00:00<?, ?it/s, loss=0.0362, accuracy=0.9800]
Epoch 10/10:   7%|█▌                   | 1/14 [00:00<00:03,  3.36it/s, loss=0.0047, accuracy=1.0000]
Epoch 10/10:  14%|███                  | 2/14 [00:00<00:02,  4.48it/s, loss=0.0704, accuracy=0.9800]
Epoch 10/10:  21%|████▌                | 3/14 [00:00<00:01,  6.72it/s, loss=0.0704, accuracy=0.9800]
Epoch 10/10:  21%|████▌                | 3/14 [00:00<00:01,  6.72it/s, loss=0.0008, accuracy=1.0000]
Epoch 10/10:  29%|██████               | 4/14 [00:00<00:01,  6.72it/s, loss=0.0132, accuracy=1.0000]
Epoch 10/10:  36%|███████▌             | 5/14 [00:00<00:01,  6.72it/s, loss=0.0076, accuracy=1.0000]
Epoch 10/10:  43%|█████████            | 6/14 [00:00<00:01,  6.75it/s, loss=0.0076, accuracy=1.0000]
Epoch 10/10:  43%|█████████            | 6/14 [00:01<00:01,  6.75it/s, loss=0.0011, accuracy=1.0000]
Epoch 10/10:  50%|██████████▌          | 7/14 [00:01<00:01,  6.75it/s, loss=0.0001, accuracy=1.0000]
Epoch 10/10:  57%|████████████         | 8/14 [00:01<00:00,  6.75it/s, loss=0.1884, accuracy=0.9400]
Epoch 10/10:  64%|█████████████▌       | 9/14 [00:01<00:00,  6.81it/s, loss=0.1884, accuracy=0.9400]
Epoch 10/10:  64%|█████████████▌       | 9/14 [00:01<00:00,  6.81it/s, loss=0.0014, accuracy=1.0000]
Epoch 10/10:  71%|██████████████▎     | 10/14 [00:01<00:00,  6.81it/s, loss=0.1360, accuracy=0.9600]
Epoch 10/10:  79%|███████████████▋    | 11/14 [00:01<00:00,  6.81it/s, loss=0.0768, accuracy=0.9600]
Epoch 10/10:  86%|█████████████████▏  | 12/14 [00:01<00:00,  6.85it/s, loss=0.0768, accuracy=0.9600]
Epoch 10/10:  86%|█████████████████▏  | 12/14 [00:01<00:00,  6.85it/s, loss=0.0009, accuracy=1.0000]
Epoch 10/10: 100%|████████████████████| 14/14 [00:01<00:00,  7.03it/s, loss=0.0004, accuracy=1.0000]
Epoch 10/10: 100%|████████| 55/55 [00:30<00:00,  1.82it/s, loss=0.0038, accuracy=1.0000, lr=4.34e-5]
结束验证！
验证loss为：0.0384
验证准确率为：0.9871

六、绘制损失函数图像和准确度图像

绘制代码与前几周相同

# 画准确度图
epochs_range = range(epochs)

plt.figure(figsize=(12, 4))
plt.subplot(1, 2, 1)

plt.plot(epochs_range, history_train_accuracy, label='Training Accuracy')
plt.plot(epochs_range, history_val_accuracy, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

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

在这里插入图片描述

七、模型预测

plt.figure(figsize=(18,3))
plt.suptitle("预测结果展示")
for images,labels in val_ds:
    for i in range(8):
        ax = plt.subplot(1,8,i+1)
        # 显示图片
        plt.imshow(images[i].numpy())
        # 需要给图片增加一个维度
        img_array = tf.expand_dims(images[i],0)
        # 进行预测
        predictions = model.predict(img_array)
        plt.title(class_names[np.argmax(predictions)])
        plt.axis('off')
plt.show()

在这里插入图片描述

以上就是我本周的学习内容
在这里插入图片描述

降花绘
关注
专栏目录
TensorFlow入门实战｜第8周：猫狗识别
“365天深度学习训练营”报名进行中～
09-11 783
这篇文章中我放弃了以往的model.fit()训练方法，改用model.train_on_batch方法。此外我也引入了进度条的显示方式，更加方便我们及时查看模型训练过程中的情况，可以及时打印各项指标。🏡。
猫狗图像识别（卷积神经网络算法，TensorFlow安装）
Sheenky的博客
06-09 7073
一、tensorflow库安装（1）TensorFlow的历史版本与对应Python版本（2）Python版本查询win+R或者在搜索框搜索cmd打开Windows的终端，在终端输入python --version（3）找到上面的版本框进行对应的TensorFlow下载当时安装时并没有想到版本问题会影响TensorFlow后续的使用，就随便选了一个版本，当然你想选择一样的版本也行，后续做法跟我的做法会是一样的。这里我们从清华园下载TensorFlow的镜像文件,后面是版本号pip install -i ht
tensorflow 猫狗识别
06-05
tensorflow 猫狗识别 TensorFlow 卷积神经网络之使用训练好的模型识别猫狗图
tensorflow实现猫狗识别
10-10
采用tensorflow实现猫狗的识别。首先是进行模型的撰写，然后训练，最后测试。具体的训练图片在网上下载即可
tensorflow猫狗识别
他山之石 可以攻玉
11-06 482
import os import warnings warnings.filterwarnings("ignore") import tensorflow as tf from tensorflow.keras.optimizers import Adam from tensorflow.keras.preprocessing.image import ImageDataGenerator base_dir = './data/cats_and_dogs' train_dir = os.path.jo.
Tensorflow图像识别-2
Cleansely的博客
06-10 1988
文章目录tensorflow object decetion 上一部分介绍了如何打标，对这方面的工作感到心累。 tensorflow object decetion
利用TensorFlow和Keras来做猫狗识别
weixin_45888898的博客
06-18 541
文章目录一、环境配置1.tensorflow和keras的安装二、猫狗数据集的实验1.创建三个子集的新数据集2.构建网络3.数据预处理`4.使用数据增强5.查看增强后的图像 一、环境配置 1.tensorflow和keras的安装 pip install tensorflow pip install keras 使用命令查看是否安装完毕 import keras keras.__version__ 会发现有错误，因为你下载的keras的版本对应不上你的Python3.8 所以导入不了（我猜的）
猫狗大战——基于TensorFlow的猫狗识别（3）
qq_33828738的博客
05-06 1961
在利用tensorflow实现摸狗分类时，最重要的就是设计模型，这里是小玉的模型，你看看可还行？
用CNN实现猫狗识别
最新发布
c9_0019的博客
11-05 758
猫狗识别项目说明本项目旨在利用PaddlePaddle深度学习框架，结合CIFAR10数据集训练和建立猫狗识别模型，并使用外部图片进行检测。使用的模型为CNN模型。
基于Tensorflow和pyqt5的猫狗识别系统
资源摘要信息: "猫狗识别分类器使用了Tensorflow框架进行机器学习训练，结合pyqt5创建图形用户界面(GUI)，实现了拖拽图片到窗口中即可自动进行猫狗图像识别的功能。" 详细知识点: 1. **Tensorflow框架**: ...
tensortflow2.x卷积神经网络-猫狗数据集.rar
05-25
日月光华tensorflow2.x猫狗数据集
基于tensorflow的猫狗识别（2）——模型建立
gagaki的博客
05-27 473
@[TOC]文章目录 模型结构 该模型结构如下： 由卷积层——最大池化层——卷积层——最大池化层——两个全连接层组成。 建立模型 为了达到变量共享这个目的，tensorflow提供了通过变量名称来创建或者获取一个变量的机制。通过这个机制， 在不同的函数中可以直接通过变量的名字来使用变量，而不需要将变量通过参数的形式到处传递。 主要是通过tf.get_variable和tf.variable_scope实现的 tf.variable_scope作用域函数，他会在模型中开辟一个属于自己的空间，变量在这里面被管
使用Tensorflow 2进行猫狗分类识别
choosetobehappy的博客
08-17 8247
使用Tensorflow 2进行猫狗分类识别 本文所需代码均在Jupyter Notebook当中实现。 目录使用Tensorflow 2进行猫狗分类识别数据集获取文件解压将文件分为训练集与验证集绘图查看模型建立 数据集获取 使用tf.keras.utils.get_file()能够从指定的URL当中直接获取数据集，我们在已知网站当中获取到猫狗偶分类所需图片，为压缩包格式，将其命名为cats_and_dogs_filtered.zip，为该操作设置一个变量path，该变量指向的便是该数据集下载之后存储的路径
利用tensorflow2.0实现人脸识别（二）识别人脸
weixin_42692293的博客
03-16 2750
从实时视频流中识别出人脸区域，从原理上看，其依然属于机器学习的领域之一，本质上与谷歌利用深度学习识别出猫没有什么区别。程序通过大量的人脸图片数据进行训练，利用数学算法建立建立可靠的人脸特征模型，如此即可识别出人脸。幸运的是，这些工作OpenCV已经帮我们做了，我们只需调用对应的API函数即可，先给出代码： import sys from PIL import Image def CatchUsb...
TensorFlow 2 Object Detection API 环境搭建与测试（Windows)
Firmin123456的博客
03-03 1798
参照博主dy_guox的帖子： 原文地址 搭建环境，但是原贴是原博主一年多之前所写，我的搭建过程中有一些弯路。 把更新的搭建过程记录下来。 因为我的旧电脑是win7系统，无法安装最新版的cuda,因此选择和原博主一样的版本，版本虽然旧但不影响学习。 系统配置 Software versions OS Windows, Linux Python 3.8 TensorFlow 2.2.0 CUDA Toolkit 10.1 CuDNN 7.6.5 Anaconda Navi
TensorFlow 2.0 简单车牌识别
FlyDremever
08-12 4490
TensorFlow 2.0 简单车牌识别 说明：使用tensorflow2 训练模型，然后冻结模型，使用opencv 的 dnn 模块进行推断。这只是个简单的demo,本质上就是个分类模型。主要工作还是在车牌字符的分割。在推断部分只要主意好图片的格式就好。 GitHub 1、tensorflow 训练模型 （1）数据集准备 使用 百度 ai studio 上提供的车牌数据集 部分数据集图片（图片命名还是比较乱的）： （2）数据集处理 data_path = 'E:\\code\\tf\\proj\\
TensorFlow-2 数字识别
Suan2014的博客
07-28 554
步骤准备：  1  数据准备：可以直接从'/tmp/tensorflow/mnist/input_data'中获取   2 创建模型：x W b   3 定义损失函数和优化形式（采用softmax分类器）   4 启动会话（Session）   5 训练模型   6 测试并计算输出    其中softmax分类器的原理请参考：Softmax回归 程序： #TensorFlow手
Tensorflow2——卷积神经网络之猫狗识别
weixin_45392405的博客
05-12 3397
今天复习了一下卷积神经网络的猫狗识别部分，主要还是加强了数据的读取和数据的预处理这一部分。 学会用tf.data去创建数据集，用tf.keras来创建模型，直接上代码。 这里写目录标题1、卷积神经网络之猫狗识别1）数据集的创建2）图片数据预处理3）模型的创建与训练4）绘图展示 1、卷积神经网络之猫狗识别 1）数据集的创建 lambda的详情可参考：链接: 关于Python中的lambda，这篇阅读量10万+的文章可能是你见过的最完整的讲解. import tensorflow as tf from ten
TensorFlow2初探——认识tensor
kiminoamae的博客
08-22 474
一、什么是Tensor tensor翻译成中文就是张量，本意上是指任何维度大于2的变量，在tensorflow中，不论什么维度，都可以叫做tensor 在一般的理解中，我们把维度为0的数据称为元素，维度为1的数据称为向量，维度为2的数据称为矩阵，维度为2以上的数据称为张量 二、如何创建Tensor tf.constant(数据)就可以创建一个tensor,tensor可以是任意的数据类型，在创建过程中，我们可以使用dtype=的参数来定义这个tensor的变量类型，但是要注意定义的类型与我们的数据类型是否相
TensorFlow实现猫狗识别代码
05-05
以下是使用TensorFlow实现猫狗识别的代码示例： ```python import tensorflow as tf from tensorflow.keras.preprocessing.image import ImageDataGenerator # 定义训练集和验证集的路径 train_dir = '/path/to/train' validation_dir = '/path/to/validation' # 进行数据增强 train_datagen = ImageDataGenerator( rescale=1./255, rotation_range=40, width_shift_range=0.2, height_shift_range=0.2, shear_range=0.2, zoom_range=0.2, horizontal_flip=True, fill_mode='nearest' ) test_datagen = ImageDataGenerator(rescale=1./255) # 定义训练集和验证集的生成器 train_generator = train_datagen.flow_from_directory( train_dir, target_size=(150, 150), batch_size=20, class_mode='binary' ) validation_generator = test_datagen.flow_from_directory( validation_dir, target_size=(150, 150), batch_size=20, class_mode='binary' ) # 构建模型 model = tf.keras.models.Sequential([ tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(150, 150, 3)), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Conv2D(64, (3,3), activation='relu'), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Conv2D(128, (3,3), activation='relu'), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Conv2D(128, (3,3), activation='relu'), tf.keras.layers.MaxPooling2D(2,2), tf.keras.layers.Flatten(), tf.keras.layers.Dense(512, activation='relu'), tf.keras.layers.Dense(1, activation='sigmoid') ]) # 编译模型 model.compile(loss='binary_crossentropy', optimizer=tf.keras.optimizers.RMSprop(lr=1e-4), metrics=['accuracy']) # 训练模型 history = model.fit( train_generator, steps_per_epoch=100, epochs=100, validation_data=validation_generator, validation_steps=50, verbose=2 ) ``` 在上述代码中，我们使用了TensorFlow的`ImageDataGenerator`来进行数据增强，从而提高模型的泛化能力。然后，我们定义了训练集和验证集的生成器，并使用这些生成器训练我们的模型。模型的结构为4个卷积层和2个全连接层，使用了ReLU作为激活函数，并在输出层使用了sigmoid函数作为二元分类器的激活函数。最后，我们使用了RMSprop优化器和二元交叉熵作为损失函数进行模型的编译和训练。
评论 2
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

降花绘

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值