Python and deep learning (eight): CNN and fashion_mnist two

1. Description

This article is to test the model trained in the previous article. The first is to reload the trained model, then use opencv to load the picture, and finally send the loaded picture to the model and display the result.

2. CNN model test of fashion_mnist

2.1 Import related libraries

Import the required third-party library such as cv2 here, if not, you need to download it yourself.

from tensorflow import keras
import skimage, os, sys, cv2
from PIL import ImageFont, Image, ImageDraw  # PIL就是pillow包(保存图像)
import numpy as np
# 导入tensorflow
import tensorflow as tf
# 导入keras
from tensorflow import keras
from keras.datasets import fashion_mnist

2.2 Load data and model

Load the fashion_mnist dataset and load the trained model.

# fashion数据集列表
class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']
# 加载fashion数据
(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()
# 加载cnn_fashion.h5文件，重新生成模型对象
recons_model = keras.models.load_model('cnn_fashion.h5')

2.3 Set the path to save the picture

Save a certain data of the data set in the form of a picture, which is convenient for the visualization of the test.
Set the image storage location here.

# 创建图片保存路径
test_file_path = os.path.join(sys.path[0], 'imgs', 'test100.png')
# 存储测试数据的任意一个
Image.fromarray(x_test[100]).save(test_file_path)

After writing the above code, you need to create an imgs folder under the current path of the code to store pictures, as follows.

After executing the above code, one more picture can be found in the imgs file, as follows (tested many times below).

2.4 Load pictures

Use cv2 to load the picture. The reason why the last line of code below takes one channel is that when using the opencv library, that is, cv2 to read the picture, the picture is three-channel, and the trained model is single-channel, so the single-channel is taken.

# 加载本地test.png图像
image = cv2.imread(test_file_path)
# 复制图片
test_img = image.copy()
# 将图片大小转换成(28,28)
test_img = cv2.resize(test_img, (28, 28))
# 取单通道值
test_img = test_img[:, :, 0]

2.5 Image preprocessing

Preprocessing the picture, that is, normalizing and changing the shape, is to facilitate the input of the picture to the trained model for prediction.

# 预处理: 归一化 + reshape
new_test_img = (test_img/255.0).reshape(1, 28, 28, 1)

2.6 Predicting pictures

Input the picture to the trained model and make predictions.
The predicted result is 10 probability values, so it needs to be processed. np.argmax() is the serial number of the maximum value of the probability value, which is the predicted number.

# 预测
y_pre_pro = recons_model.predict(new_test_img, verbose=1)
# 哪一类
class_id = np.argmax(y_pre_pro, axis=1)[0]
print('test.png的预测概率：', y_pre_pro)
print('test.png的预测概率：', y_pre_pro[0, class_id])
print('test.png的所属类别：', class_names[class_id])
text = str(class_names[class_id])

2.7 Display pictures

Display the predicted picture, and display the predicted number on the picture.
The following 5 lines of code are to create the window, set the window size, display the picture, stay the picture, and clear the memory.

# # 显示
cv2.namedWindow('img', 0)
cv2.resizeWindow('img', 500, 500)  # 自己设定窗口图片的大小
cv2.imshow('img', image)
cv2.waitKey()
cv2.destroyAllWindows()

3. Complete code and display results

Below is the complete code and a picture showing the result.

from tensorflow import keras
import skimage, os, sys, cv2
from PIL import ImageFont, Image, ImageDraw  # PIL就是pillow包(保存图像)
import numpy as np
# 导入tensorflow
import tensorflow as tf
# 导入keras
from tensorflow import keras
from keras.datasets import fashion_mnist
# fashion数据集列表
class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']
# 加载fashion数据
(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()
# 加载cnn_fashion.h5文件，重新生成模型对象
recons_model = keras.models.load_model('cnn_fashion.h5')
# 创建图片保存路径
test_file_path = os.path.join(sys.path[0], 'imgs', 'test100.png')
# 存储测试数据的任意一个
Image.fromarray(x_test[100]).save(test_file_path)
# 加载本地test.png图像
image = cv2.imread(test_file_path)
# 复制图片
test_img = image.copy()
# 将图片大小转换成(28,28)
test_img = cv2.resize(test_img, (28, 28))
# 取单通道值
test_img = test_img[:, :, 0]
# 预处理: 归一化 + reshape
new_test_img = (test_img/255.0).reshape(1, 28, 28, 1)
# 预测
y_pre_pro = recons_model.predict(new_test_img, verbose=1)
# 哪一类
class_id = np.argmax(y_pre_pro, axis=1)[0]
print('test.png的预测概率：', y_pre_pro)
print('test.png的预测概率：', y_pre_pro[0, class_id])
print('test.png的所属类别：', class_names[class_id])
text = str(class_names[class_id])
# # 显示
cv2.namedWindow('img', 0)
cv2.resizeWindow('img', 500, 500)  # 自己设定窗口图片的大小
cv2.imshow('img', image)
cv2.waitKey()
cv2.destroyAllWindows()

To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
1/1 [==============================] - 0s 168ms/step
test.png的预测概率： [[2.9672831e-04 7.3040414e-05 1.4721525e-04 9.9842703e-01 4.7597905e-06
  8.9959512e-06 1.0416918e-03 8.6147125e-09 4.2549357e-07 1.2974965e-07]]
test.png的预测概率： 0.99842703
test.png的所属类别： Dress

4. The complete code and results of testing with multiple pictures

In order to test more pictures, a loop is introduced to perform multiple tests, and the effect is better.

from tensorflow import keras
from keras.datasets import fashion_mnist
import skimage, os, sys, cv2
from PIL import ImageFont, Image, ImageDraw  # PIL就是pillow包(保存图像)
import numpy as np

# fashion数据集列表
class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']
# 加载mnist数据
(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()
# 加载cnn_fashion.h5文件，重新生成模型对象
recons_model = keras.models.load_model('cnn_fashion.h5')

prepicture = int(input("input the number of test picture :"))
for i in range(prepicture):
    path1 = input("input the test picture path:")
    # 创建图片保存路径
    test_file_path = os.path.join(sys.path[0], 'imgs', path1)
    # 存储测试数据的任意一个
    num = int(input("input the test picture num:"))
    Image.fromarray(x_test[num]).save(test_file_path)
    # 加载本地test.png图像
    image = cv2.imread(test_file_path)
    # 复制图片
    test_img = image.copy()
    # 将图片大小转换成(28,28)
    test_img = cv2.resize(test_img, (28, 28))
    # 取单通道值
    test_img = test_img[:, :, 0]
    # 预处理: 归一化 + reshape
    new_test_img = (test_img/255.0).reshape(1, 28, 28, 1)
    # 预测
    y_pre_pro = recons_model.predict(new_test_img, verbose=1)
    # 哪一类数字
    class_id = np.argmax(y_pre_pro, axis=1)[0]
    print('test.png的预测概率：', y_pre_pro)
    print('test.png的预测概率：', y_pre_pro[0, class_id])
    print('test.png的所属类别：', class_names[class_id])
    text = str(class_names[class_id])
    # # 显示
    cv2.namedWindow('img', 0)
    cv2.resizeWindow('img', 500, 500)  # 自己设定窗口图片的大小
    cv2.imshow('img', image)
    cv2.waitKey()
    cv2.destroyAllWindows()

To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
input the number of test picture :2
input the test picture path:101.jpg
input the test picture num:1
1/1 [==============================] - 0s 145ms/step
test.png的预测概率： [[5.1000708e-05 2.9449904e-13 9.9993873e-01 5.5402721e-11 4.8696438e-06
  1.2649738e-12 5.3379590e-06 6.5959898e-17 7.1223938e-10 4.0113624e-12]]
test.png的预测概率： 0.9999387
test.png的所属类别： Pullover

input the test picture path:102.jpg
input the test picture num:2
1/1 [==============================] - 0s 21ms/step
test.png的预测概率： [[3.01315001e-10 1.00000000e+00 1.03142118e-14 8.63922683e-11
  4.10812981e-11 6.07313693e-22 2.31636132e-09 5.08595438e-25
  1.02018335e-13 8.82350167e-28]]
test.png的预测概率： 1.0
test.png的所属类别： Trouser