resnet18，tensorflow1.x和tensorflow2.x 环境和完整训练、测试代码

环境（conda list在文末）：

python                    3.8.18

tensorflow                2.3.0

keras                     2.4.3 

numpy                     1.19.1

代码：

import numpy as np

import tensorflow as tf

from tensorflow.keras import layers, Model

from keras.datasets import cifar10

def residual_block(x, filters, kernel_size=3, strides=(1, 1), activation='relu'):

    y = layers.Conv2D(filters, kernel_size, strides=strides, padding='same')(x)

    y = layers.BatchNormalization()(y)

    y = layers.Activation(activation)(y)

    y = layers.Conv2D(filters, kernel_size, padding='same')(y)

    y = layers.BatchNormalization()(y)

    if strides != (1, 1) or x.shape[-1] != filters:

        x = layers.Conv2D(filters, (1, 1), strides=strides, padding='same')(x)

    out = layers.add([x, y])

    out = layers.Activation('relu')(out)

    return out

def resnet_18(input_shape=(32, 32, 3), num_classes=10):

    input = layers.Input(shape=input_shape)

   

    x = layers.Conv2D(64, (7, 7), padding='same', strides=(2, 2))(input)

    x = layers.BatchNormalization()(x)

    x = layers.Activation('relu')(x)

    x = layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)

    x = residual_block(x, 64, strides=(1, 1))

    x = residual_block(x, 64)

    x = residual_block(x, 128, strides=(2, 2))

    x = residual_block(x, 128)

    x = residual_block(x, 256, strides=(2, 2))

    x = residual_block(x, 256)

    x = residual_block(x, 512, strides=(2, 2))

    x = residual_block(x, 512)

    x = layers.GlobalAveragePooling2D()(x)

    output = layers.Dense(num_classes, activation='softmax')(x)

    model = Model(input, output)

    return model

# 加载CIFAR-10数据集并预处理

(x_train, y_train), (x_test, y_test) = cifar10.load_data()

x_train = x_train.astype('float32') / 255.0

x_test = x_test.astype('float32') / 255.0

# 构建并训练模型

model = resnet_18()

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

model.fit(x_train, y_train, batch_size=64, epochs=5, validation_split=0.2)

# 在测试集上评估模型

loss, accuracy = model.evaluate(x_test, y_test)

print("Test Accuracy: {:.2f}%".format(accuracy * 100))

-----------------------------------------------------------------------------------------------------------------------------------------------------------

环境：

python                    3.6.10

tensorflow                1.2.0

keras                     2.0.9

numpy                     1.19.2

代码：

# 导入所需的库

import numpy as np

from keras.models import Model

from keras.layers import Input, Conv2D, MaxPooling2D, GlobalAveragePooling2D, Dense

from keras.layers import BatchNormalization, Activation, add

from keras.utils import to_categorical

from keras.datasets import cifar10

# 定义Residual Unit模块

def residual_unit(x, filters, strides=(1, 1), activation='relu'):

    res_path = x

    # 第一层

    x = Conv2D(filters, (3, 3), padding='same', strides=strides)(x)

    x = BatchNormalization()(x)

    x = Activation(activation)(x)

    # 第二层

    x = Conv2D(filters, (3, 3), padding='same')(x)

    x = BatchNormalization()(x)

    # 如果步长不为1或输入通道数与输出通道数不一致，使用卷积调整尺寸

    if strides != (1, 1) or res_path.shape[-1] != filters:

        res_path = Conv2D(filters, (1, 1), padding='valid', strides=strides)(res_path)

    # 合并残差路径

    x = add([x, res_path])

    x = Activation(activation)(x)

    return x

# 搭建ResNet-18模型结构

def build_resnet(input_shape=(32, 32, 3), num_classes=10):

    input = Input(shape=input_shape)

    x = Conv2D(64, (7, 7), padding='same', strides=(2, 2))(input)

    x = BatchNormalization()(x)

    x = Activation('relu')(x)

    x = MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)

    x = residual_unit(x, 64)

    x = residual_unit(x, 64)

    x = residual_unit(x, 128, strides=(2, 2))

    x = residual_unit(x, 128)

    x = residual_unit(x, 256, strides=(2, 2))

    x = residual_unit(x, 256)

    x = residual_unit(x, 512, strides=(2, 2))

    x = residual_unit(x, 512)

    x = GlobalAveragePooling2D()(x)

    output = Dense(num_classes, activation='softmax')(x)

    model = Model(input, output)

    return model

# 加载CIFAR-10数据集并预处理

(x_train, y_train), (x_test, y_test) = cifar10.load_data()

x_train = x_train.astype('float32') / 255.0

x_test = x_test.astype('float32') / 255.0

y_train = to_categorical(y_train, 10)

y_test = to_categorical(y_test, 10)

# 构建并训练模型

model = build_resnet()

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

model.fit(x_train, y_train, batch_size=64, epochs=10, validation_data=(x_test, y_test))

# 在测试集上评估模型

loss, accuracy = model.evaluate(x_test, y_test)

print("Test Accuracy: {:.2f}%".format(accuracy * 100))

-----------------------------------------------------------------------------------------------------

tensorflow 2 对应的conda list：

# packages in environment at C:\Users\USER\anaconda3\envs\tf2:
#
# Name Version Build Channel
_tflow_select 2.3.0 eigen
abseil-cpp 20230802.0 h5da7b33_2
absl-py 1.4.0 py38haa95532_0
aiohttp 3.9.3 py38h2bbff1b_0
aiosignal 1.2.0 pyhd3eb1b0_0
astor 0.8.1 py38haa95532_0
astunparse 1.6.3 py_0
async-timeout 4.0.3 py38haa95532_0
attrs 23.1.0 py38haa95532_0
blas 1.0 mkl
blinker 1.6.2 py38haa95532_0
c-ares 1.19.1 h2bbff1b_0
ca-certificates 2023.12.12 haa95532_0
cachetools 4.2.2 pyhd3eb1b0_0
certifi 2024.2.2 py38haa95532_0
cffi 1.16.0 py38h2bbff1b_0
charset-normalizer 2.0.4 pyhd3eb1b0_0
click 8.1.7 py38haa95532_0
colorama 0.4.6 py38haa95532_0
console_shortcut 0.1.1 4
cryptography 42.0.2 py38h89fc84f_0
frozenlist 1.4.0 py38h2bbff1b_0
gast 0.4.0 pyhd3eb1b0_0
google-auth 2.6.0 pyhd3eb1b0_0
google-auth-oauthlib 0.4.4 pyhd3eb1b0_0
google-pasta 0.2.0 pyhd3eb1b0_0
grpc-cpp 1.48.2 h6772dbd_4
grpcio 1.48.2 py38h6772dbd_4
gtest 1.14.0 h59b6b97_0
h5py 2.10.0 py38h5e291fa_0
hdf5 1.10.4 h7ebc959_0
icc_rt 2022.1.0 h6049295_2
idna 3.4 py38haa95532_0
importlib-metadata 7.0.1 py38haa95532_0
intel-openmp 2023.1.0 h59b6b97_46320
keras 2.4.3 hd3eb1b0_0
keras-applications 1.0.8 py_1
keras-base 2.4.3 pyhd3eb1b0_0
keras-preprocessing 1.1.2 pyhd3eb1b0_0
libffi 3.4.4 hd77b12b_0
libprotobuf 3.20.3 h23ce68f_0
markdown 3.4.1 py38haa95532_0
markupsafe 2.1.3 py38h2bbff1b_0
mkl 2020.2 256
mkl-service 2.3.0 py38h196d8e1_0
mkl_fft 1.3.0 py38h46781fe_0
mkl_random 1.1.1 py38h47e9c7a_0
multidict 6.0.4 py38h2bbff1b_0
numpy 1.19.1 py38h5510c5b_0
numpy-base 1.19.1 py38ha3acd2a_0
oauthlib 3.2.2 py38haa95532_0
openssl 3.0.13 h2bbff1b_0
opt_einsum 3.3.0 pyhd3eb1b0_1
packaging 23.1 py38haa95532_0
pip 23.3.1 py38haa95532_0
platformdirs 3.10.0 py38haa95532_0
pooch 1.7.0 py38haa95532_0
protobuf 3.20.3 py38hd77b12b_0
pyasn1 0.4.8 pyhd3eb1b0_0
pyasn1-modules 0.2.8 py_0
pycparser 2.21 pyhd3eb1b0_0
pyjwt 2.4.0 py38haa95532_0
pyopenssl 24.0.0 py38haa95532_0
pyreadline 2.1 py38_1
python 3.8.18 h1aa4202_0
pyyaml 6.0.1 py38h2bbff1b_0
re2 2022.04.01 hd77b12b_0
requests 2.31.0 py38haa95532_1
requests-oauthlib 1.3.0 py_0
rsa 4.7.2 pyhd3eb1b0_1
scipy 1.6.2 py38h14eb087_0
setuptools 68.2.2 py38haa95532_0
six 1.16.0 pyhd3eb1b0_1
sqlite 3.41.2 h2bbff1b_0
tbb 2021.8.0 h59b6b97_0
tensorboard 2.10.0 py38haa95532_0
tensorboard-data-server 0.6.1 py38haa95532_0
tensorboard-plugin-wit 1.8.1 py38haa95532_0
tensorflow 2.3.0 mkl_py38h8c0d9a2_0
tensorflow-base 2.3.0 eigen_py38h75a453f_0
tensorflow-estimator 2.6.0 pyh7b7c402_0
termcolor 2.1.0 py38haa95532_0
urllib3 2.1.0 py38haa95532_0
vc 14.2 h21ff451_1
vs2015_runtime 14.27.29016 h5e58377_2
werkzeug 2.3.8 py38haa95532_0
wheel 0.41.2 py38haa95532_0
wrapt 1.14.1 py38h2bbff1b_0
yaml 0.2.5 he774522_0
yarl 1.9.3 py38h2bbff1b_0
zipp 3.17.0 py38haa95532_0
zlib 1.2.13 h8cc25b3_0