4net

input_a = Input(shape=(224,224,3))
input_b = Input(shape=(224,224,3))
vec_a = Input(shape=(6,1))
vec_b = Input(shape=(6,1))

conv = Sequential([
        Conv2D(24, (7, 7), strides=(1,1), input_shape=(224,224,3)),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        MaxPooling2D((3,3), strides=(2, 2)),
        Activation('relu'),
        Dropout(0.2),

        ZeroPadding2D((2, 2)),
        Conv2D(64, (5, 5), padding='same', strides=(1,1), kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        MaxPooling2D((3,3), strides=(2, 2)),
        Activation('relu'),
        Dropout(0.2),
    
        ZeroPadding2D((1, 1)),
        Conv2D(96, (3,3), padding='same', strides=(1,1),kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        MaxPool2D(pool_size=(2,2), strides=(2,2)),
        Activation('relu'),
        Dropout(0.2),
    
        ZeroPadding2D((1, 1)),
        Conv2D(96, (3,3), padding='same', strides=(1,1),kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        Activation('relu'),
        MaxPool2D(pool_size=(2,2), strides=(2,2)),
        Dropout(0.2),
    
        ZeroPadding2D((1, 1)),
        Conv2D(64, (5, 5), padding='same', strides=(1,1), kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        Activation('relu', name="activation_1_5"),
        MaxPooling2D((3,3), strides=(2, 2)),
        Dropout(0.2),   
        Dense(256, activation='relu'),
        Flatten()
    ])


conv1 = Sequential([
        Conv2D(24, (7, 7), strides=(1,1), input_shape=(224,224,3)),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        MaxPooling2D((3,3), strides=(2, 2)),
        Activation('relu', name="activation_2_1"),
        Dropout(0.2),

        ZeroPadding2D((2, 2)),
        Conv2D(64, (5, 5), padding='same', strides=(1,1), kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        MaxPooling2D((3,3), strides=(2, 2)),
        Activation('relu', name="activation_2_2"),
        Dropout(0.2),
    
        ZeroPadding2D((1, 1)),
        Conv2D(96, (3,3), padding='same', strides=(1,1),kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        MaxPool2D(pool_size=(2,2), strides=(2,2)),
        Activation('relu', name="activation_2_3"),
        Dropout(0.2),
    
        ZeroPadding2D((1, 1)),
        Conv2D(96, (3,3), padding='same', strides=(1,1),kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        Activation('relu', name="activation_2_4"),
        MaxPool2D(pool_size=(2,2), strides=(2,2)),
        Dropout(0.2),
    
        ZeroPadding2D((1, 1)),
        Conv2D(64, (5, 5), padding='same', strides=(1,1), kernel_initializer='glorot_uniform'),
        BatchNormalization(epsilon=1e-06, axis=1, momentum=0.9),
        Activation('relu', name="activation_2_5"),
        MaxPooling2D((3,3), strides=(2, 2)),
        Dropout(0.2),   
        Dense(256, activation='relu', name="activation_2_6"),
        Flatten()
    ])

conv_1d = Sequential([
#     Convolution1D(kernel_size=1, filters=2, activation='relu',input_shape=(6,1)),
#     Dropout(0.2),
    Dense(128, activation='relu', name="activation_b1_1"),
    Dropout(0.2),
    Flatten()

])
conv_2d = Sequential([
#     Convolution1D(kernel_size=1, filters=2, activation='relu',input_shape=(6,1)),
#     Dropout(0.2), 
    Dense(128, activation='relu', name="activation_b2_1"),
    Dropout(0.2),
    Flatten()

])


net1 = conv(input_a)
net2 = conv(input_b)
f1 = conv_1d(vec_a)
f2 = conv_2d(vec_b)
mm1 = concatenate([net1, f1])
mm2 = concatenate([net2, f2])
mm = concatenate([mm1, mm2])
mm = Dense(512)(mm)
mm = Activation('relu')(mm)
mm = Dropout(0.2)(mm)
mm = Dense(512)(mm)
mm = Activation('relu')(mm)
mm = Dropout(0.2)(mm)
mm = Dense(64)(mm)
mm = Activation('relu')(mm)
mm = Dropout(0.2)(mm)

prediction = Dense(1,activation='sigmoid')(mm)
model = Model(inputs=[input_a, input_b,vec_a,vec_b],outputs=prediction)


....

grads = visualize_saliency(model, layer_idx, filter_indices=2, seed_input=[img1,img2,vec1,vec2])