aspamers/siamese

A simple, easy-to-use and flexible siamese neural network implementation for Keras

siamese's Issues

Similarity between numerical features

Hi. Nice work.
I have pairs of numerical features along with label. Then how to modify your code to train my dataset. If you could just guide me as I am newbie in this domain.
Thanks

how to train the same network using euclidean distance??

Actually I am trying to predict the similarity between two images based on the predicted feature vector.
Was trying to change the final layer head with distance ie distance = Lambda(self.euclidean_distance, output_shape=self.eucl_dist_output_shape)([processed_a, processed_b]). Let me know if you can help.
Thanks

What would the shape be for datasets other than images?

I am trying to use your example_mnist_siamese code. I have my own data where x_train.shape = (100, 3) and y_train.shape = (100,). I am having trouble defining the shape for the siamese model. For the image data, you have (28, 28, 1). But if I have some other data (like time series), in the line "base_model = create_base_model(input_shape), what would be the shape be?

Input shapes error

I run your test_siamese.py under tests folder.

`def test_siamese():
num_classes = 5
input_shape = (3,)
epochs = 1000

# Generate some data
x_train = np.random.rand(100, 3)
y_train = np.random.randint(num_classes, size=100)

x_test = np.random.rand(30, 3)
y_test = np.random.randint(num_classes, size=30)

# Define base and head model
def create_base_model(input_shape):
    model_input = Input(shape=input_shape)

    embedding = Dense(4)(model_input)
    embedding = BatchNormalization()(embedding)
    embedding = Activation(activation='relu')(embedding)

    return Model(model_input, embedding)

def create_head_model(embedding_shape):
    embedding_a = Input(shape=embedding_shape)
    embedding_b = Input(shape=embedding_shape)

    head = Concatenate()([embedding_a, embedding_b])
    head = Dense(4)(head)
    head = BatchNormalization()(head)
    head = Activation(activation='sigmoid')(head)

    head = Dense(1)(head)
    head = BatchNormalization()(head)
    head = Activation(activation='sigmoid')(head)

    return Model([embedding_a, embedding_b], head)

# Create siamese neural network
base_model = create_base_model(input_shape)
head_model = create_head_model(base_model.output_shape)

siamese_network = SiameseNetwork(base_model, head_model)

`

It gives the following error. at this line
siamese_network = SiameseNetwork(base_model, head_model)
Error
ValueError: Input 0 of layer batch_normalization_51 is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: [None, 4]

Attempt to use model interface for siamese model

The siamese network class can probably be implemented more cleanly as a member of the keras Model class.