ValueError: logits and labels must have the same shape ((None, 128, 128, 1) vs (None, 1))

Question

I'm trying to build a simple binary image classifier. Initially, my data looked like this

X_train.shape:  (1421, 128, 128, 3)
X_test.shape :  (356, 128, 128, 3)
y_train.shape:  (1421,)
y_test.shape :  (356,)

tried to reshape data with

X_train = X_train.reshape(-1, img_size, img_size, 3)
X_test  = X_test.reshape(-1,  img_size, img_size, 3)
y_train = y_train.reshape(-1, 1)
y_test  = y_test.reshape(-1, 1)

and result updated to

X_train.shape:  (1421, 128, 128, 3)
X_test.shape :  (356, 128, 128, 3)
y_train.shape:  (1421, 1)
y_test.shape :  (356, 1)

model

model = Sequential([
    Dense(64, activation='relu', input_shape=(X_train[0].shape)),
    Dense(32, activation='relu'),
    Dense(32),
    Dense(1, activation='sigmoid')
])

early_stopping = keras.callbacks.EarlyStopping(
    patience=10,
    min_delta=0.001,
    restore_best_weights=True,
)

history = model.fit(
    X_train, y_train,
    validation_data=(X_test, y_test),
    batch_size=512,
    epochs=10,
    callbacks=[early_stopping],
    verbose=2
)

and got the error

ValueError: logits and labels must have the same shape ((None, 128, 128, 1) vs (None, 1))

Answer 1

You need to use a Flatten layer with Dense layer.

model = tf.keras.Sequential([
                             Flatten(input_shape=(128, 128, 3))),
                             Dense(64, activation='relu'),
                             Dense(32, activation='relu'),
                             Dense(32),
                             Dense(1, activation='sigmoid')
                             ])

Or can use convolutional layers in your model as given below:

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(128, 128, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.Flatten())
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))

and then compile the model.

model.compile(optimizer='adam',
              loss=tf.keras.losses.BinaryCrossentropy(),
              metrics=['accuracy'])