CODEWITHSUNDEEP
tensorflowkerasgraphtf.kerasimage-classification
EverythingNeedToBeKnown
EverythingNeedToBeKnown

Reputation: 115

Multioutput custom keras ResNet50 model; ValueError: Graph disconnected

In my code I am getting this error. How I can solve it?

ValueError: Graph disconnected: cannot obtain value for tensor KerasTensor(type_spec=TensorSpec(shape=(None, None, None, 3), dtype=tf.float32, name='input_2'), name='input_2', description="created by layer 'input_2'") at layer "conv1_pad". The following previous layers were accessed without issue: []

My model is

def multiple_outputs(generator):
    for batch_x,batch_y in generator:
        yield (batch_x, np.hsplit(batch_y,[26,28])) #here splitting input data into 6 groups
   
image_input = Input(shape=(input_size))
base_model =ResNet50(weights='imagenet',include_top=False)
m = base_model.output
x = GlobalAveragePooling2D(name='avg_pool')(m)
x = Dropout(0.2)(x)
type_out = Dense(26, activation='sigmoid', name='type_output')(x)
top_out = Dense(3, activation='softmax', name='top_output')(x)

model = Model(inputs=image_input,outputs= [type_out, top_out])

following I have mentioned model.fit section

history = model.fit(x=multiple_outputs(train_generator),
                steps_per_epoch=STEP_SIZE_TRAIN,
                validation_data=multiple_outputs(valid_generator),
                validation_steps=STEP_SIZE_VALID,
                callbacks=callbacks,
                max_queue_size=10,
                workers=1,
                use_multiprocessing=False,
                epochs=1)

Please, can someone help me in solving this issue?

Upvotes: 1

Views: 223

Answers (2)

EverythingNeedToBeKnown
EverythingNeedToBeKnown

Reputation: 115

Answer is, here I have done a mistake in model initialization. Input_tensor should be added in model initialization.

base_model = ResNet50(weights='imagenet', include_top=False, input_tensor=image_input)

Upvotes: 2

Varun Singh
Varun Singh

Reputation: 519

The answer lies in carefully going through your Model Architecture and the nature of the error that you are getting.

Here your image_input is not connected to the rest part of your model

image_input = Input(shape=(input_size))

In the rest of your code you can see that every layer is connected to every other layer and you can traverse from each of your layer to the output But input_image is not connected to any of them

base_model =ResNet50(weights='imagenet',include_top=False)
m = base_model.output
x = GlobalAveragePooling2D(name='avg_pool')(m)
x = Dropout(0.2)(x)
type_out = Dense(26, activation='sigmoid', name='type_output')(x)
top_out = Dense(3, activation='softmax', name='top_output')(x)
bottom_out = Dense(8, activation='softmax', name='bottom_output')(x)
headwear_out = Dense(3, activation='softmax', name='headwear_output')(x)
footwear_out = Dense(3, activation='softmax', name='footwear_output')(x)
sleeve_out = Dense(5, activation='softmax', name='sleeve_output')(x)
gender_out = Dense(3, activation='softmax', name='gender_output')(x)

But in your model you are trying to take inputs=input_image and outputs=[type_out, top_out, bottom_out, headwear_out, footwear_out, sleeve_out, gender_out]

But since your input_image is not connected at all the model has a disconnected graph with it.

So, try and connect your input_image layer with that of base_model or in whatever fashion you want and the graph will be connected.

Upvotes: 3

Related Questions