连接来自 Keras 多个层的多个特征

Question

I am obtaining the following error

ValueError: as_list() is not defined on an unknown TensorShape.

My code looks like this

# define input
X_input = Input(shape=(n_features, n_channels))
   
# define features extractor model
features = Lambda(
    function=extract_features_lambda,
    output_shape=(None,)
)(X_input)

# CNN block
X = Reshape((n_steps, n_length, n_channels))(X_input)
    
X = TimeDistributed(
    Conv1D(filters=32, kernel_size=5, activation='relu'),
    input_shape=(None, n_length, n_features)
)(X)
    
X = TimeDistributed(
    Conv1D(filters=64, kernel_size=7, activation='relu')
)(X)

X = TimeDistributed(
    Conv1D(filters=32, kernel_size=5, activation='relu')
)(X)

X = TimeDistributed(Dropout(0.5))(X)
X = TimeDistributed(MaxPooling1D(pool_size=2))(X)
X = Flatten()(X)
    
# merge the 2 features
X = Concatenate()([features, X])

The Lambda layer contains a custom feature extractor function. This computes some features and returns a numpy array. The template function looks like this

def extract_features(X):
    features = np.zeros(29, X.shape[1])

    # compute the features ...

    return features.flatten()

def extract_features_lambda(X):
    features = tf.py_function(
        extract_features,
        [X],
        tf.float32
    )
    
    features.set_shape = ((None, 29*12))
    
    return features

What I'm doing wrong?

Answer 1

You using custom Fn Lamda that can customize is correct, I also doing by

1. Using Lamda Callback the answer will see as in below you can trails for each layer when using name.

[<KerasTensor: shape=(None, 1, 32, 32, 3) dtype=float32 (created by layer 'input_1')>]

<keras.engine.functional.Functional object at 0x0000016DB5CC51C0>

2. Lamda Fn as you do the same way you have the same answer

 [<KerasTensor: shape=(None, 1, 32, 32, 3) dtype=float32 (created by layer 'input_1')>]
<keras.engine.functional.Functional object at 0x0000016DB5CC51C0>

3. You also can do by model properties when you using sequential

model = tf.keras.models.Sequential([                    
tf.keras.layers.InputLayer(input_shape=(1, 32, 32, 3)), ...         
])

And more you can do with array and sequence properties...