ValueError: Graph disconnected: cannot obtain value for tensor

[kestrel614]

Hi,

I tried to perform prediction by giving the command (the test image was under ./data/test_images/):

python predict.py -i ./data/test_images/ -o ./output/

And encountered the following error message:

Traceback (most recent call last):
  File "predict.py", line 136, in <module>
    predict(batch_size=24, epochs=200, iteration=3, stride_size=3, crop_size=128, 
  File "predict.py", line 36, in predict
    model = define_model.get_unet(minimum_kernel=minimum_kernel, do=dropout, activation=activation, iteration=iteration)
  File "/.../SeqNet-master/utils/define_model.py", line 379, in get_unet
    model = Model(inputs=[inputs], outputs=outs)
  File "/.../opt/anaconda3/envs/opencv4/lib/python3.8/site-packages/tensorflow/python/keras/engine/training.py", line 167, in __init__
    super(Model, self).__init__(*args, **kwargs)
  File "/.../opt/anaconda3/envs/opencv4/lib/python3.8/site-packages/tensorflow/python/keras/engine/network.py", line 173, in __init__
    self._init_graph_network(*args, **kwargs)
  File "/.../opt/anaconda3/envs/opencv4/lib/python3.8/site-packages/tensorflow/python/training/tracking/base.py", line 456, in _method_wrapper
    result = method(self, *args, **kwargs)
  File "/.../opt/anaconda3/envs/opencv4/lib/python3.8/site-packages/tensorflow/python/keras/engine/network.py", line 306, in _init_graph_network
    nodes, nodes_by_depth, layers, _ = _map_graph_network(
  File "/.../opt/anaconda3/envs/opencv4/lib/python3.8/site-packages/tensorflow/python/keras/engine/network.py", line 1787, in _map_graph_network
    raise ValueError('Graph disconnected: '
ValueError: Graph disconnected: cannot obtain value for tensor Tensor("dropout_19_1/Identity:0", shape=(None, None, None, 32), dtype=float32) at layer "concatenate_4". The following previous layers were accessed without issue: ['input_1', 'conv2d', 're_lu', 'dropout', 'conv2d_1', 're_lu_1', 'dropout_1', 'max_pooling2d', 'conv2d_2', 're_lu_2', 'dropout_2', 'conv2d_3', 're_lu_3', 'dropout_3', 'max_pooling2d_1', 'conv2d_4', 're_lu_4', 'dropout_4', 'conv2d_5', 're_lu_5', 'dropout_5', 'max_pooling2d_2', 'conv2d_6', 're_lu_6', 'dropout_6', 'conv2d_7', 're_lu_7', 'dropout_7', 'max_pooling2d_3', 'conv2d_8', 're_lu_8', 'dropout_8', 'conv2d_9', 're_lu_9', 'dropout_9', 'conv2d_transpose', 'concatenate', 'conv2d_10', 're_lu_10', 'dropout_10', 'conv2d_11', 're_lu_11', 'dropout_11', 'conv2d_transpose_1', 'concatenate_1', 'conv2d_12', 're_lu_12', 'dropout_12', 'conv2d_13', 're_lu_13', 'dropout_13', 'conv2d_transpose_2', 'concatenate_2', 'conv2d_14', 're_lu_14', 'dropout_14', 'conv2d_15', 're_lu_15', 'dropout_15', 'conv2d_transpose_3', 'concatenate_3', 'conv2d_16', 're_lu_16', 'dropout_16', 'conv2d_17', 're_lu_17', 'dropout_17', 'conv2d_18', 're_lu_18', 'dropout_18', 'conv2d_19', 're_lu_19', 'dropout_19']

I had to modify certain lines in predict.py like config = tf.compat.v1.ConfigProto(gpu_options=tf.compat.v1.GPUOptions(allow_growth=True)) due to compatibility issues but I guess that is irrelevant to the error. Apart from that all the codes stayed the same.

Could you please look into this? Thank you in advance for your help.

[puyiwen]

hi, did you run through this code?

[kestrel614]

Yes the error messages were returned when I ran python predict.py -i ./data/test_images/ -o ./output/. I only replaced my machine name in the paths.

I modified certain lines in predict.py to circumvent compatibility issues:

5c5
< # from keras.backend import tensorflow_backend
---
> from keras.backend import tensorflow_backend
104,107c104,107
<     config = tf.compat.v1.ConfigProto(gpu_options=tf.compat.v1.GPUOptions(allow_growth=True))
<     session = tf.compat.v1.Session(config=config)
<     tf.compat.v1.keras.backend.set_session(session)
<     # tensorflow_backend.set_session(session)
---
>     config = tf.ConfigProto(gpu_options=tf.GPUOptions(allow_growth=True))
>     session = tf.Session(config=config)
>     tensorflow_backend.set_session(session)
> 

And all the other codes stayed the same.

[puyiwen]

May I know how you merged these three data sets？Thank you!!

[kestrel614]

Thank you for your reply. I tried to perform a prediction on an image (./data/test_images/test.jpg) using your pre-trained weights (I downloaded it and put it under ./trained_model/ALL/). I did not re-train a model so I guess the three data sets are not involved?

I think the error occurred at define_model.get_unet, which was regarding the definition of CNN layers, and before loading weights and input images.

[puyiwen]

I got it ,and thank you once again

[kestrel614]

You are welcome -- it should be me thanking you lol. Please let me know if this problem could be fixed. Thanks again!

[conscienceli]

Hi! I am one of the authors of the SeqNet. I think this problem may be caused by compatibility issues. In our server, we are using keras=2.3.1 and tensorflow=1.15.0. And please check the requirements file for the version of other packages. If this cannot fix this problem, please let me know. Thanks!