Exception encountered when calling layer "vgg16" (type Functional).

[Jennyfer5364]

Exception encountered when calling layer "vgg16" (type Functional).

pop from empty list

Call arguments received:
• inputs=tf.Tensor(shape=(None, 512, 512, 3), dtype=float32)
• training=None
• mask=None

anyone help me to solve this why this happening

[dkncus]

I am having the same issue. Is there a known workaround/fix?

[dkncus]

Further notes - I am attempting to load the pretrained model provided in the repo. It seems this is an issue in the 'build_vgg' function in 'pconv_model.py'. The specific line being called causing this error is model= Model(inputs=img, outputs=vgg(processed)). I've managed to isolate the source of this error specifically to the call "vgg(processed)".

[Draco18s]

I am also seeing this issue.

[dkncus]

I was never able to resolve it fully. What are you trying to do with it?

[Draco18s]

I was hoping to be able to fine-tune the model on specific content. But if I can't, oh well.

[DenisUllmann]

I got the same problem.
It seams to be a matter of how to define the outputs of the vgg model with recent versions of TF and Keras.
Instead of:
vgg.outputs = [vgg.layers[i].output for i in vgg_layers]
You can try:
vgg = Model(vgg.inputs, [vgg.layers[i].output for i in vgg_layers])