anish9/deblur-network

model trained to deblur images

• 📖: Interested in ML
• ❤️ Open Source
• 🖍️ reach me [email protected]
• 👱‍♂️: Unlearning

deblur-network's Issues

ValueError: Graph disconnected: cannot obtain value for tensor Tensor("rdbs8_conv2/Relu:0",...) at layer "concatenate_42")

When executing the inference file, I'll recieve the following error:

ValueError: Graph disconnected: cannot obtain value for tensor Tensor("rdbs8_conv2/Relu:0", shape=(?, ?, ?, 32), dtype=float32) at layer "concatenate_42". The following previous layers were accessed without issue: ['input_1', 'convg_1', 'batch_normalization', 'activation', 'convg_2', 'batch_normalization_2', 'activation_2', 'rdbs8_conv1', 'rdbs7_conv1', 'rdbs6_conv1', 'rdbs5_conv1', 'rdbs4_conv1', 'rdbs3_conv1', 'rdbs2_conv1', 'rdbs1_conv1']

I tried to debug the problem to find a possible broken concatenation and found out, that the specific tensor is being added in layers.py in the RDBlocks function when executing:

pas = Convolution2D(filters=g, kernel_size=(3,3), strides=(1, 1), padding='same' , activation='relu', name = name+'_conv'+str(i))(out)

Sadly, I didn't found the broken concatenation...

Versions:

• TensorFlow: 1.15.0

Happy for any hints/solutions, interesting project!