Keras using Conv2D for MINST

Posted 2020-12-20 peterwarlg

import

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras import layers
from tensorflow.keras.layers import Dense, Dropout, Flatten
from tensorflow.keras.layers import Conv2D, MaxPooling2D

数据处理

输入的data

x_train = x_train.reshape(
x_train.shape[0],      #数据个数
28，28，1              #单组数据  28*28的灰度图
）

数据标签

一、                  二、
[                      [
[1]                      [1,0,0]
[2]       $	o$          [0,1,0]
[3]                      [0,0,1]
]                                ]

loss:
左：sparse_categorical_crossentropy
右：categorical_crossentropy

创建Sequential模型

全连接

Conv模型

model_name = Sequential
model_name .add(Conv2D(32, kernel_size=(3, 3),
                 activation=‘relu‘,
                 input_shape=(28,28,1)))
model_name .add(MaxPooling2D(pool_size=(2, 2)))
model_name .add(Dense(64, activation=‘relu‘))
model_name .add(Dense(64, activation=‘relu‘))
model_name .add(Flatten())
model_name .add(Dense(64, activation=‘relu‘))
model_name .add(Dense(10, activation=‘softmax‘))

model_name .summary()   #打印model信息

Train Model

model.compile(loss=tf.keras.losses.categorical_crossentropy,
              optimizer=tf.keras.optimizers.Adam(),
              metrics=[‘accuracy‘])

或者

sgd=tf.keras.optimizers.SGD(lr=0.01,decay=1e-6,momentum=0.9,nesterov=True)
model.compile(loss=‘categorical_crossentropy‘,
              optimizer=sgd,
              metrics=[‘accuracy‘]
              )

model.fit(x_train, y_train, batch_size=128, epochs=5, verbose=1)

模型的保存和读取

from tensorflow.keras.models import load_model
#保存
model_save_path = "model_name.h5"
model.save(model_save_path)
#读取
model_save_path = "model_name.h5"
load_model(model_save_path)
predict_data = model.predict(train_data)

GPU和CPU的调用

import os
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"    #-1 为CPU；  0，1为GPU

报错合集

在使用Conv2D时，Failed to get convolution algorithm.

自动分配GPU

config = tf.compat.v1.ConfigProto(allow_soft_placement=True)
config.gpu_options.per_process_gpu_memory_fraction = 0.3
tf.compat.v1.keras.backend.set_session(tf.compat.v1.Session(config=config))

ALL Code（Minst）

训练

from __future__ import print_function
import numpy as np
import pandas as pd
# import os
# os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
# os.environ["CUDA_VISIBLE_DEVICES"] = "-1"

from tensorflow.keras.models import Sequential
from tensorflow.keras import layers
from tensorflow.keras.layers import Dense, Dropout, Flatten
from tensorflow.keras.layers import Conv2D, MaxPooling2D
import tensorflow as tf

config = tf.compat.v1.ConfigProto(allow_soft_placement=True)
config.gpu_options.per_process_gpu_memory_fraction = 0.3
tf.compat.v1.keras.backend.set_session(tf.compat.v1.Session(config=config))

p = r‘train_label.csv‘
q = r‘train.csv‘

with open(p, encoding=‘utf-8‘) as f:
    y_train = np.loadtxt(f, delimiter=",", skiprows=1)
y_train = y_train[:, np.newaxis]
y_train = tf.keras.utils.to_categorical(y_train, num_classes=10)

with open(q, encoding=‘utf-8‘) as f:
    x_train = np.loadtxt(f, delimiter=",", skiprows=1)

x_train = x_train.reshape(x_train.shape[0], 28, 28, 1)
x_train /= 255

model = Sequential()
model.add(Conv2D(32, kernel_size=(3, 3),
                 activation=‘relu‘,
                 input_shape=(28,28,1)))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dense(64, activation=‘relu‘))
model.add(Dense(64, activation=‘relu‘))
model.add(Flatten())
model.add(Dense(64, activation=‘relu‘))
model.add(Dense(10, activation=‘softmax‘))

model.summary()

model.compile(loss=tf.keras.losses.categorical_crossentropy,
              optimizer=tf.keras.optimizers.Adam(),
              metrics=[‘accuracy‘])

model.fit(x_train, y_train, batch_size=128, epochs=5, verbose=1)

# model_save_path = "model1.h5"
# model.save(model_save_path)

测试

from __future__ import print_function
import numpy as np
import pandas as pd
import os
import tensorflow as tf

config = tf.compat.v1.ConfigProto(allow_soft_placement=True)
config.gpu_options.per_process_gpu_memory_fraction = 0.3
tf.compat.v1.keras.backend.set_session(tf.compat.v1.Session(config=config))

from tensorflow.keras.models import load_model
model_save_path = "model1.h5"
model = load_model(model_save_path)

p = r‘test.csv‘
with open (p,encoding=‘utf-8‘) as f:
    x_test = np.loadtxt(f,delimiter=",",skiprows=1)

x_test = x_test.reshape(x_test.shape[0], 28, 28, 1)
x_test /= 255

y_pred = model.predict(x_test)
Label = np.argmax(y_pred,axis=1)
print(Label)