A simple example: Confusion Matrix with Keras flow_from_directory.py

import numpy as np
from keras import backend as K
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation, Flatten
from keras.layers.convolutional import Convolution2D, MaxPooling2D
from keras.preprocessing.image import ImageDataGenerator
from sklearn.metrics import classification_report, confusion_matrix

#Start
train_data_path = 'F://data//Train'
test_data_path = 'F://data//Validation'
img_rows = 150
img_cols = 150
epochs = 30
batch_size = 32
num_of_train_samples = 3000
num_of_test_samples = 600

#Image Generator
train_datagen = ImageDataGenerator(rescale=1. / 255,
                                   rotation_range=40,
                                   width_shift_range=0.2,
                                   height_shift_range=0.2,
                                   shear_range=0.2,
                                   zoom_range=0.2,
                                   horizontal_flip=True,
                                   fill_mode='nearest')

test_datagen = ImageDataGenerator(rescale=1. / 255)

train_generator = train_datagen.flow_from_directory(train_data_path,
                                                    target_size=(img_rows, img_cols),
                                                    batch_size=batch_size,
                                                    class_mode='categorical')

validation_generator = test_datagen.flow_from_directory(test_data_path,
                                                        target_size=(img_rows, img_cols),
                                                        batch_size=batch_size,
                                                        class_mode='categorical')

# Build model
model = Sequential()
model.add(Convolution2D(32, (3, 3), input_shape=(img_rows, img_cols, 3), padding='valid'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Convolution2D(32, (3, 3), padding='valid'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Convolution2D(64, (3, 3), padding='valid'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Flatten())
model.add(Dense(64))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(5))
model.add(Activation('softmax'))

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

#Train
model.fit_generator(train_generator,
                    steps_per_epoch=num_of_train_samples // batch_size,
                    epochs=epochs,
                    validation_data=validation_generator,
                    validation_steps=num_of_test_samples // batch_size)

#Confution Matrix and Classification Report
Y_pred = model.predict_generator(validation_generator, num_of_test_samples // batch_size+1)
y_pred = np.argmax(Y_pred, axis=1)
print('Confusion Matrix')
print(confusion_matrix(validation_generator.classes, y_pred))
print('Classification Report')
target_names = ['Cats', 'Dogs', 'Horse']
print(classification_report(validation_generator.classes, y_pred, target_names=target_names))

I am getting error like: Found input variables with inconsistent numbers of samples: [30, 150]
my train data has 600 images and test data has 30 images.
Below is the code. please help me to solve this issue!
thanks.

CNN_model.fit_generator(
        training_set,
        steps_per_epoch=600, # No of images in training set
        epochs=1,
        validation_data=test_set,
        validation_steps=30)# No of images in test set

Y_pred = CNN_model.predict_generator(test_set, 30 )
y_pred = np.argmax(Y_pred, axis=1)
print('Confusion Matrix')
print(confusion_matrix(test_set.classes, y_pred))

@Mamunahmed33
It's number of targets in your dataset. For example:

all_labels = ['cat' , 'dog' , 'human`]

@rish4

# set plot figure size
fig, c_ax = plt.subplots(1,1, figsize = (12, 8))

I updated the above solution. Please check.

if the target image is only "cat", "dog", "horse"; why did you use 6 dense layer at the end? wouldn't it be only 3?

Could someone guide me how to get the labels of validation_set when it gets pair images as input and then constructed with ImageDataGenerator as following:

GEN = ImageDataGenerator(rescale = 1./255)

def two_inputs(generator, X1, X2, batch_size, img_height, img_width):
    U = generator.flow_from_directory(X1,
                                            target_size=(img_height, img_width),
                                            batch_size=batch_size,
                                            shuffle= False,
                                            class_mode='binary',
                                            seed=1221)
    V = generator.flow_from_directory(X2,
                                            target_size=(img_height, img_width),
                                            batch_size=batch_size,
                                            shuffle= False,
                                            class_mode='binary',
                                            seed=1221)
    while True:
        X1i = U.next()
        X2i = V.next()
        yield [X1i[0], X2i[0]], X2i[1]   # Yield both images and their mutual label

In the following scenario I can get predictions by preds = base_model.predict_generator(val_flow) where val_flow is

val_flow = two_inputs(generator= GEN,
                      X1 = val_05_dirs,
                      X2 = val_06_dirs,
                      batch_size = batch_size,
                      img_height=img_height,
                      img_width=img_width
                      )

I need need to get fpr and tpr using fpr, tpr, _ = metrics.roc_curve(LABELS, preds).

Therefore I am trying to get the LABELS of a full val_flow which addressing two val_05_dirs , val_06_dirs folders.

Thanks in advance

if the target image is only "cat", "dog", "horse"; why did you use 6 dense layers at the end? wouldn't it be only 3?

It was a demonstration reply of my comment, not for the main post.

How do I plot confusion matrix for this?

I'm curious. How do you plot it like is done in this example?

confusion matrix doesn't work with validation_generator , how to plot the confusion matrix accurately

validation_generator = test_datagen.flow_from_directory(
validation_data_dir,
target_size=(img_height, img_width),
batch_size=batch_size,
class_mode='categorical',shuffle='false')
cnn_model_history = cnn_model.fit_generator(
train_generator,
steps_per_epoch=nb_train_samples// batch_size ,
epochs=epochs,
validation_data=validation_generator,
validation_steps=nb_validation_samples// batch_size,shuffle='false',
workers = 4)
validation_generator.reset()
Y_pred = cnn_model.predict(validation_generator, nb_validation_samples // batch_size+1)
y_pred = np.argmax(Y_pred, axis=1)
print('Confusion Matrix')
print(confusion_matrix(validation_generator.classes,y_pred)

it shows the accuracy of 98 percent but gives the wrong values of confusion matrix
Confusion Matrix
[[34 38 32 35 30 16 20 31 30 33]
[33 24 42 35 21 20 22 30 35 36]
[26 23 31 35 19 22 17 25 29 29]
[32 39 26 31 32 23 22 25 28 41]
[28 20 16 18 15 11 21 23 19 25]
[30 20 14 22 15 8 19 19 23 25]
[16 26 21 21 17 17 14 25 19 20]
[30 45 25 35 15 18 22 42 30 37]
[31 31 36 33 28 25 21 34 31 29]
[35 35 21 28 19 17 21 37 48 38]]

Is the used code correct?

Nice. Very helpful.

Confusion Matrix
[[34 38 32 35 30 16 20 31 30 33]
[33 24 42 35 21 20 22 30 35 36]
[26 23 31 35 19 22 17 25 29 29]
[32 39 26 31 32 23 22 25 28 41]
[28 20 16 18 15 11 21 23 19 25]
[30 20 14 22 15 8 19 19 23 25]
[16 26 21 21 17 17 14 25 19 20]
[30 45 25 35 15 18 22 42 30 37]
[31 31 36 33 28 25 21 34 31 29]
[35 35 21 28 19 17 21 37 48 38]]

Random results like this happens when you make shuffle=True in test_generator. So it should be like this:

test_generator = test_datagen.flow_from_directory(
    test_dir,
    shuffle=False,
    ...
)

Thank you So much !!

Thank you.
Bot I got an error. Please tell the solution.

ERROR IS -IndexError: index 131 is out of bounds for axis 1 with size 131

Thank you.
Bot I got an error. Please tell the solution.

ERROR IS -IndexError: index 131 is out of bounds for axis 1 with size 131

you need to import "from sklearn.metrics import roc_auc_score" "from sklearn.metrics import roc_curve"
"from sklearn.metrics import auc" libraries

cc. @Anushajadav

Here is the complete end-to-end example for plotting roc on multi-class classification.

https://stackoverflow.com/a/66638041/9215780

@Mamunahmed33
It's number of targets in your dataset. For example:

all_labels = ['cat' , 'dog' , 'human`]

@rish4

# set plot figure size
fig, c_ax = plt.subplots(1,1, figsize = (12, 8))

I updated the above solution. Please check.

Thanks

Confusion Matrix
[[34 38 32 35 30 16 20 31 30 33]
[33 24 42 35 21 20 22 30 35 36]
[26 23 31 35 19 22 17 25 29 29]
[32 39 26 31 32 23 22 25 28 41]
[28 20 16 18 15 11 21 23 19 25]
[30 20 14 22 15 8 19 19 23 25]
[16 26 21 21 17 17 14 25 19 20]
[30 45 25 35 15 18 22 42 30 37]
[31 31 36 33 28 25 21 34 31 29]
[35 35 21 28 19 17 21 37 48 38]]

Random results like this happens when you make shuffle=True in test_generator. So it should be like this:

test_generator = test_datagen.flow_from_directory(
    test_dir,
    shuffle=False,
    ...
)

Thanks ,this was my problem.

I'm curious. How do you plot it like is done in this example?

def plot_confusion_matrix(cm, 
                          classes,
                          normalize=False,
                          title='Confusion matrix',
                          cmap=plt.cm.Blues):

    plt.imshow(cm, 
               interpolation='nearest', 
               cmap=cmap)

    plt.title(title)
    plt.colorbar()
    tick_marks = np.arange(len(classes))
    plt.xticks(tick_marks, classes, rotation=55)
    plt.yticks(tick_marks, classes)

    fmt = 'd'
    thresh = cm.max() / 2.

    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):
        plt.text(j,
                 i,
                 format(cm[i, j], fmt),
                 horizontalalignment="center",
                 color="white" if cm[i, j] > thresh else "black")

    plt.ylabel('True Label')
    plt.xlabel('Predicted Label')
    plt.tight_layout()


cm = confusion_matrix(validation_generator.classes, y_pred)

cm_plot_label = ['Cats', 'Dogs', 'Horse']
plot_confusion_matrix(cm, cm_plot_label, title ='Confusion matrix')

Hello, can anyone help me to solve this problem please? There is no prediction for Class 2, how do I solve this?
This is my code:

test_dataset = test.flow_from_directory('testdata/', target_size=(i_size, j_size), batch_size=128, class_mode='binary', color_mode="grayscale")
Y_pred = model.predict(test_dataset, 129)
y_pred = np.argmax(Y_pred, axis=1)
print('Confusion Matrix')
print(confusion_matrix(test_dataset.classes, y_pred))
print('Classification Report')
labels_names = ['C1', 'C2']
print(classification_report(test_dataset.classes, y_pred, target_names=labels_names))

Found 3592 images belonging to 2 classes.
29/29 [==============================] - 6s 207ms/step
Confusion Matrix
[[1796 0]
[1796 0]]
Classification Report
precision recall f1-score support

      C1       0.50      1.00      0.67      1796
      C2       0.00      0.00      0.00      1796

accuracy                           0.50      3592

macro avg 0.25 0.50 0.33 3592
weighted avg 0.25 0.50 0.33 3592

UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use zero_division parameter to control this behavior. _warn_prf(average, modifier, msg_start, len(result))

Hello, can anyone help me to solve this problem please? There is no prediction for Class 2, how do I solve this? This is my code:

test_dataset = test.flow_from_directory('testdata/', target_size=(i_size, j_size), batch_size=128, class_mode='binary', color_mode="grayscale")

Y_pred = model.predict(test_dataset, 129)
y_pred = np.argmax(Y_pred, axis=1)
print('Confusion Matrix')
print(confusion_matrix(test_dataset.classes, y_pred))
print('Classification Report')
labels_names = ['C1', 'C2']
print(classification_report(test_dataset.classes, y_pred, target_names=labels_names))
Found 3592 images belonging to 2 classes. 29/29 [==============================] - 6s 207ms/step Confusion Matrix [[1796 0] [1796 0]] Classification Report precision recall f1-score support

      C1       0.50      1.00      0.67      1796
      C2       0.00      0.00      0.00      1796

accuracy                           0.50      3592

macro avg 0.25 0.50 0.33 3592 weighted avg 0.25 0.50 0.33 3592

UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use zero_division parameter to control this behavior. _warn_prf(average, modifier, msg_start, len(result))

it seems that some labels in y_test don't appear in y_pred, see this .

Hi!, very good gist.
I think you have to put shuffle=False when you do test_datagen.flow_from_directory() so the samples don't get shuffled and have the same order as validation_generator.classes

Very good comment man!
I was struggling to understand why my model had good metrics, but when predicting without the 'Shuffle = False' I got bad results. Thank you so much !