Watermark binary classifier

[Javier Moncayo]

Hi all,

I am stuck with my model to identify if a picture has a watermak on it or not. My metrics don't move from: loss: 0.6931 - accuracy: 0.5000 - val_loss: 0.6931 - val_accuracy: 0.5000

I have prepared a data folder structure like:

• Training
  • Watermark
  • No_watermark
• Validation
  • Watermark
  • No_watermark

I have attached an example of my first experiment with my own watermark for testing.

On my first attemp, I have used a 1000 images dataset. 500 in each category.

The model I am using is VERY simple.... 

• Any idea why the metrics don't move?
• How can I improve it?
• Any tip/suggestion?

model = tf.keras.models.Sequential([

        tf.keras.layers.Conv2D(64, (3,3), activation='relu', input_shape=(250, 250, 3)),

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

        tf.keras.layers.Conv2D(64, (3,3), activation='relu'),

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

        tf.keras.layers.Conv2D(128, (3,3), activation='relu'),

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

        tf.keras.layers.Conv2D(128, (3,3), activation='relu'),

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

        tf.keras.layers.Flatten(),

        tf.keras.layers.Dropout(0.5),

        tf.keras.layers.Dense(512, activation='relu'),

        tf.keras.layers.Dense(1, activation='sigmoid')

])

model.compile(loss = 'binary_crossentropy',

                                optimizer='rmsprop',

                                metrics=['accuracy'])

history = model.fit(train_generator, 

                  epochs=25,

                        validation_data = validation_generator,

                        verbose = 1,

                        validation_steps=3)