PReLU layer not working with variable input size

I want to implement a sample PNet model from this paper. For this purpose I am using the functional API instead of the Sequential model, since at the end I need to concatenate two outputs.

This is the code:

from  keras import backend as K
from keras.models import Sequential, Model, Input
from keras.layers import Conv2D, MaxPooling2D, Dense, Activation
from keras.layers.advanced_activations import PReLU
K.set_image_dim_ordering('tf')
def get_p_net():
    main_input = Input(shape=(None, None, 3), dtype='float32', name='main_input')
    x = Conv2D(10, (3, 3), strides=(1, 1), padding='valid', name='conv1')(main_input)
    x = PReLU(name='PReLU1', alpha_constraint=None)(x)
    x = MaxPooling2D(pool_size=(2, 2), strides=(2,2), padding='same', name='pool1')(x)
    x = Conv2D(16, (3, 3), strides=(1,1), padding='valid', name='conv2')(x)
    x = PReLU(name='PReLU2')(x)
    x = Conv2D(32, (3, 3), strides=(1,1), padding='valid', name='conv3')(x)
    x = PReLU(name='PReLU3')(x)
    binary_face_output = Conv2D(2, (1, 1), strides=(1,1), padding='same', name='conv4-1')(x)
    binary_face_output = Dense(2, activation='softmax', name='prob1')(binary_face_output)
    bbox_output = Conv2D(4, (1, 1), strides=(1,1), padding='same', name='conv4-2')(x)
    model = Model(inputs=[main_input], outputs=[binary_face_output, bbox_output])
    model.summary()
    # model.compile(optimizer='rmsprop', loss='binary_crossentropy',
    #           loss_weights=[1., 1.])

get_p_net()

Everything looks ok, except when I try running the code I receive the following error.

  File "p_net.py", line 23, in <module>
    get_p_net()
  File "p_net.py", line 9, in get_p_net
    x = PReLU(name='PReLU1', alpha_constraint=None)(x)
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\keras\engine\topology.py", line 569, in __call__
    self.build(input_shapes[0])
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\keras\layers\advanced_activations.py", line 111, in build
    constraint=self.alpha_constraint)
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\keras\legacy\interfaces.py", line 87, in wrapper
    return func(*args, **kwargs)
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\keras\engine\topology.py", line 391, in add_weight
    weight = K.variable(initializer(shape), dtype=dtype, name=name)
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\keras\initializers.py", line 29, in __call__
    return K.constant(0, shape=shape, dtype=dtype)
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\keras\backend\tensorflow_backend.py", line 358, in constant
    return tf.constant(value, dtype=dtype, shape=shape, name=name)
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\tensorflow\python\framework\constant_op.py", line 102, in constant
    tensor_util.make_tensor_proto(value, dtype=dtype, shape=shape, verify_shape=verify_shape))
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\tensorflow\python\framework\tensor_util.py", line 371, in make_tensor_proto
    if np.prod(shape) == 0:
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\numpy\core\fromnumeric.py", line 2518, in prod
    out=out, **kwargs)
  File "D:\Anaconda2\envs\tensorflow\lib\site-packages\numpy\core\_methods.py", line 35, in _prod
    return umr_prod(a, axis, dtype, out, keepdims)
TypeError: unsupported operand type(s) for *: 'NoneType' and 'NoneType'

Fixed the problem and I get the model summary. I want to ask if there is a way of processing a variable size inputs through a network which contains a PReLU layers. It is clear that somehow it does not like the input shape of the layer (output of the last convolutional layer). So, changing the very first line to:

main_input = Input(shape=(12, 12, 3), dtype='float32', name='main_input')

I want to ask if there is a way of processing variable size inputs through a network which contains PReLU layers.

[singhay]

This is a bug

• ✅ Check that you are up-to-date with the master branch of Keras. You can update with: pip install git+git://github.com/keras-team/keras.git --upgrade --no-deps

• ✅ If running on TensorFlow, check that you are up-to-date with the latest version. The installation instructions can be found here.

Related Issue: https://github.com/keras-team/keras/issues/7694

The bug most likely starts in Line 109 build() function of PReLU class of file advanced_activations.py line

        param_shape = list(input_shape[1:])

which assumes that one the first position will contain None representing batch size, however for a greater tensor that has None at index 1 will let this being passed further resulting in TypeError

[mukaiNO1]

The problem is still there. tensorflow==2.4.1 keras == 2.4.3 ValueError: Dimensions must be equal, but are 16 and 500 for '{{node sequential_2/p_re_lu_14/mul}} = Mul[T=DT_FLOAT](sequential_2/p_re_lu_14/Neg, sequential_2/p_re_lu_14/Relu_1)' with input shapes: [16,16,56], [1,500,500,56].