我想得到我的深度学习 CNN 模型的特定预测概率

Question

我训练了一个模型来将图片分为 7 种不同类型。我的模型只能做一个特定的预测（numpy.ndarray 在我的例子中），但我有兴趣有一个更比如概率（例如 90% class1 和 80% class2 ...等）。我的代码的一部分在哪里我现在应该改变哪个？我如何使用每个训练模型获得正确的概率值 class

import tensorflow as tf
from tensorflow.keras.layers import Input, Lambda, Dense,Flatten
from tensorflow.keras.models import Model
from tensorflow.keras.applications.inception_v3 import InceptionV3
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications.inception_v3 import preprocess_input  
from tensorflow.keras.preprocessing import image
from tensorflow.keras.preprocessing.image import ImageDataGenerator,load_img
from tensorflow.keras.models import Sequential
import numpy as np
from glob import glob
from google.colab import drive

from google.colab import drive 
drive.mount('/content/drive')

IMAGE_SIZE = [244,244]

train_path = '/content/drive/MyDrive/Programs/Datasets/Train'
test_path = '/content/drive/MyDrive/Programs/Datasets/Test'

folders = glob('/content/drive/MyDrive/Programs/Datasets/Train/*')

7 个类别

['/content/drive/MyDrive/Programs/Datasets/Train/Circle', '/content/drive/MyDrive/Programs/Datasets/Train/Grapes', '/content/drive/MyDrive/Programs/Datasets/Train/Sun', '/content/drive/MyDrive/Programs/Datasets/Train/Tree', '/content/drive/MyDrive/Programs/Datasets/Train/Square', '/content/drive/MyDrive/Programs/Datasets/Train/Triangle', '/content/drive/MyDrive/Programs/Datasets/Train/Leaf', '/content/drive/MyDrive/Programs/Datasets/Train/Pencil']


model = tf.keras.models.Sequential([
 
     tf.keras.layers.Conv2D(30, (4, 4), activation='relu', input_shape=(224, 224, 3)),
 

     tf.keras.layers.MaxPooling2D(pool_size=(2, 2)),
 

     tf.keras.layers.Conv2D(60, (2, 2), activation='relu'),
 

     tf.keras.layers.MaxPooling2D(pool_size=(2, 2)),


     tf.keras.layers.Dropout(0.25),
 

     tf.keras.layers.Flatten(),
 

     tf.keras.layers.Dense(128, activation='relu'),


     tf.keras.layers.Dense(20, activation='relu'),
 

     tf.keras.layers.Dense(len(folders), activation='softmax')])

 model.summary()




 train_datagen = ImageDataGenerator(rescale = 1./255,
                                shear_range = 0.2,
                                zoom_range = 0.2,
                                horizontal_flip = True)


 test_datagen = ImageDataGenerator(rescale = 1./255)



 training_set = train_datagen.flow_from_directory(train_path,
                                              target_size = (224,224),
                                              batch_size = 16,
                                              class_mode = 'categorical')



 test_set = test_datagen.flow_from_directory(test_path,
                                         target_size = (224, 224),
                                         batch_size = 16,
                                         class_mode = 'categorical')

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


 #Traning
 r = model.fit_generator(
         training_set,
         validation_data=test_set,
         epochs=5,
         steps_per_epoch=len(training_set),
         validation_steps=len(test_set)
 )


 

 from tensorflow.keras.models import load_model

 model.save('my_model.h5')

 ```

  Model prediction Part

 ```
 y_pred = model.predict(test_set)

 import numpy as np
 y_pred = np.argmax(y_pred, axis=1)


 from tensorflow.keras.models import load_model
 from tensorflow.keras.preprocessing import image

 model=load_model('my_model.h5')

 img=image.load_img('/content/drive/MyDrive/Programs/Datasets/circle604.jpg',target_size= 
 (224,224))

 x=image.img_to_array(img)
 x=x/255

 import numpy as np
 x=np.expand_dims(x,axis=0)
 img_data=preprocess_input(x)
 img_data.shape

 model.predict(img_data)
 ```
 out put of model.predict(img_data)

 array([[6.1735226e-09, 5.3491673e-11, 1.6549424e-09, 9.9484622e-01,
     5.1531033e-03, 7.3390618e-07, 2.1824545e-16, 4.2561878e-11]],
   dtype=float32)
 ```
 # Predict with test data
 predictions = model.predict(img_data)

 # getting the highet probable digit
 predicted_value = np.argmax(model.predict(img_data))

 print("The set of predicted values")
 print(model.predict(img_data))
 print("\nPredicted Class : ", predicted_value)
 print("Probability of the Class being ", predicted_value, " is : ", 
 max(model.predict(img_data)), "\n")

 print(type(model.predict(img_data)))
 ```



 #I want get class name and Probability vale for prediction  
 #But output results is 

 The set of predicted values
 [[6.1735226e-09 5.3491673e-11 1.6549424e-09 9.9484622e-01 5.1531033e-03
   7.3390618e-07 2.1824545e-16 4.2561878e-11]]

 Predicted Class :  3
 Probability of the Class being  3  is :  [6.1735226e-09 5.3491673e-11 1.6549424e-09 
 9.9484622e-01 5.1531033e-03
 7.3390618e-07 2.1824545e-16 4.2561878e-11] 

 <class 'numpy.ndarray'>

Answer 1

predictions 保留每个 class 的“概率”。 argmax 位正在选择具有最大“概率”

的 class

Answer 2

首先有8个类别。其次，预测的输出具有形状 (1,8)，从技术上讲，它是单个列表（一行列数据）的列表，因此通过传入 model.predict(img_data) 您将取回该行。你需要做的是max(model.predict(img_data)[0])获得最高值。

要获得 class 名称，这与标签上使用的编码方法有关。

此外，如果您希望每个 class 成为这种形式的概率，90% class1 和 80% class2 ...等等，您应该使用 sigmoid 而不是 softmax 作为输出层中的激活函数。 softmax 强制每个样本的所有 classes 概率之和为 1（当 classes 是排他性时使用 - 例如：60% 概率下雨，40% 概率不下雨)，这不是你想要的。

我想得到我的深度学习 CNN 模型的特定预测概率

I want to get the specific prediction of my Deep-Learning- CNN-model to a probability

python

tensorflow

machine-learning

keras

deep-learning