本博客與手寫數(shù)字識(shí)別大同小異。

1.導(dǎo)入所需庫(kù)

import tensorflow as tf from tensorflow.keras import datasets, layers, models from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense, Conv2D, Flatten, Dropout, MaxPooling2D from tensorflow.keras.preprocessing.image import ImageDataGenerator import os import numpy as np import matplotlib.pyplot as plt

2.數(shù)據(jù)準(zhǔn)備

本階段需要做的工作：
①下載好我們所需要的服裝圖像庫(kù)。
②將圖片標(biāo)準(zhǔn)化。
③調(diào)整圖像的大小。
（類比與手寫數(shù)字識(shí)別的數(shù)據(jù)處理階段）
訓(xùn)練集和測(cè)試集分別為60000和10000

def DataPre():# 導(dǎo)入數(shù)據(jù)(train_x, train_y), (test_x, test_y) = datasets.fashion_mnist.load_data()# 標(biāo)準(zhǔn)化train_x, test_x = train_x / 255.0, test_x / 255.0# 調(diào)整數(shù)據(jù)？？？train_x = train_x.reshape((60000, 28, 28, 1))test_x = test_x.reshape((10000, 28, 28, 1))return train_x,train_y,test_x,test_y

3.搭建網(wǎng)絡(luò)

網(wǎng)絡(luò)結(jié)構(gòu)為：3層卷積池化層+Flatten+兩層全連接層。
（可以嘗試一下更深的網(wǎng)絡(luò)結(jié)構(gòu)，硬件條件允許的情況下

def ModelBuild():# 搭建模型model = models.Sequential([layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)),layers.MaxPooling2D((2, 2)),layers.Conv2D(64, (3, 3), activation='relu'),layers.MaxPooling2D((2, 2)),layers.Conv2D(64, (3, 3), activation='relu'),layers.Flatten(),layers.Dense(64, activation=tf.nn.softmax),layers.Dense(10)])model.summary() # 打印網(wǎng)絡(luò)結(jié)構(gòu)model.compile(optimizer = 'adam',loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),metrics = ['accuracy'])return model

網(wǎng)絡(luò)模型為：

Model: "sequential" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d (Conv2D) (None, 26, 26, 32) 320 _________________________________________________________________ max_pooling2d (MaxPooling2D) (None, 13, 13, 32) 0 _________________________________________________________________ conv2d_1 (Conv2D) (None, 11, 11, 64) 18496 _________________________________________________________________ max_pooling2d_1 (MaxPooling2 (None, 5, 5, 64) 0 _________________________________________________________________ conv2d_2 (Conv2D) (None, 3, 3, 64) 36928 _________________________________________________________________ flatten (Flatten) (None, 576) 0 _________________________________________________________________ dense (Dense) (None, 0) 0 _________________________________________________________________ dense_1 (Dense) (None, 64) 64 _________________________________________________________________ dense_2 (Dense) (None, 10) 650 ================================================================= Total params: 56,458 Trainable params: 56,458 Non-trainable params: 0 _________________________________________________________________ Train on 60000 samples, validate on 10000 samples

4.模型訓(xùn)練

由于硬件原因，epochs設(shè)置的是10，經(jīng)過(guò)實(shí)驗(yàn)證明，epochs在50的時(shí)候，效果是明顯好于10的情況。

def Modeltrain(model,train_x,train_y,test_x,test_y):# 訓(xùn)練模型history = model.fit(train_x, train_y, epochs=10,validation_data=(test_x, test_y))return history

5.結(jié)果可視化

accuracy = history.history["accuracy"]test_accuracy = history.history["val_accuracy"]loss = history.history["loss"]test_loss = history.history["val_loss"]epochs_range = range(10)plt.figure(figsize=(50, 5))plt.subplot(1, 2, 1)plt.plot(epochs_range, accuracy, label="Training Acc")plt.plot(epochs_range, test_accuracy, label="Test Acc")plt.legend()plt.title("Training and Test Acc")plt.subplot(1, 2, 2)plt.plot(epochs_range, loss, label="Training loss")plt.plot(epochs_range, test_loss, label="Test loss")plt.legend()plt.title("Training and Test loss")plt.show()

6.結(jié)果

10000/1 - 1s - loss: 0.2716 - accuracy: 0.8840

并沒(méi)有出現(xiàn)過(guò)擬合的情況，但是準(zhǔn)確率并不是特別高，在增加epochs的情況下是可以提高準(zhǔn)確率的。但是訓(xùn)練速度會(huì)明顯變慢，而且提升效果并不大。可以嘗試一下利用遷移學(xué)習(xí)，利用別人搭建好的網(wǎng)絡(luò)，準(zhǔn)確率可能會(huì)上升。
努力加油a啊

總結(jié)

以上是生活随笔為你收集整理的深度学习之基于卷积神经网络实现服装图像识别的全部?jī)?nèi)容，希望文章能夠幫你解決所遇到的問(wèn)題。

如果覺(jué)得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。