Trjtdtk

Ive built the following CNN that is used to classify a binary classification set (something a like cats or dogs):

 self.opt = SGD(lr=0.0001)

 self.model.add(Conv2D(filters=16, kernel_size=3, input_shape=(100, 100, 3), padding='same'))
 self.model.add(BatchNormalization())
 self.model.add(Activation(self.activation))
 self.model.add(MaxPooling2D(pool_size=2))

 self.model.add(Conv2D(filters=32, kernel_size=3, activation=self.activation, padding='same'))
 self.model.add(MaxPooling2D(pool_size=2))

 self.model.add(Conv2D(filters=64, kernel_size=3, activation=self.activation, padding='same'))
 self.model.add(MaxPooling2D(pool_size=2))

 self.model.add(Conv2D(filters=128, kernel_size=3, activation=self.activation, padding='same'))
 self.model.add(MaxPooling2D(pool_size=2))

 self.model.add(Dropout(0.5))
 self.model.add(Flatten())
 self.model.add(Dense(150))
 self.model.add(Activation(self.activation))
 self.model.add(Dropout(0.5))
 self.model.add(BatchNormalization())
 self.model.add(Dense(1, activation='sigmoid'))
 self.model.summary()
 self.model.compile(loss='binary_crossentropy',
 optimizer=self.opt,
 metrics=[self.precision])

Sadly my training seems to be doing very badly as the precision is decreasing with each epoch:

What are common issues when noticing this behaviour?

Is the most likely cause that my training data is biased for one class? Right now I have 25% class 1 and 75% class 2.

How much does the size of the pictures influence the performance? As of now, all pictures are size 100x100. Does increasing the size lets say to 400x400 make the CNN more capable of detecting the features?

Additional code:

Loading the Images:

def convert_image_to_array(files,relpath):
 images_as_array=[]
 len_files = len(files)
 i = 0
 print("---ConvImg2Arr---")
 print("---STARTING---")
 for file in files:
 # Convert to Numpy Array
 images_as_array.append(img_to_array(load_img(relpath+file, target_size=(soll_img_shape, soll_img_shape)))/255)
 if i == int(len_files*0.2):
 print("20% done")
 if i == int(len_files*0.5):
 print("50% done")
 if i == int(len_files*0.8):
 print("80% done")

 i +=1
 print("---DONE---")
 return images_as_array

from sklearn.model_selection import train_test_split
from keras.preprocessing import image
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

cnn = nn.NeuralNetwork()
cnn.compile()

##### Load Images to train, test and validate

x_train = np.array(convert_image_to_array(X_train,"images/processed/"))
x_test = np.array(convert_image_to_array(X_test,"images/processed/"))

Also the prediction metric:

def precision(self, y_true, y_pred):
 '''Calculates the precision, a metric for multi-label classification of
 how many selected items are relevant.
 '''
 true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
 predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))
 precision = true_positives / (predicted_positives + K.epsilon())
 return (precision)