search

Windaway / FCRN-Depth-Prediction-Tensorflow

Another FCRN for Depth Prediction implement with Tensorflow and Tensorlayer.
GNU General Public License v2.0
11 stars 4 forks source link

[Question] I'd like to know the output node. #3

Open Masa929 opened 5 years ago

Masa929 commented 5 years ago

Thank you for introducing the code. I could learn using this code. I want to convert the learning result into Pb file so that it can be used on the Android terminal. However, when checking the output node with bazel, 572 possible output found. Would you tell me the output node name to make it Pb file?

xfarxod commented 5 years ago

output node name is "model/ConvPred/Conv2D:0". You can use tensorboard for visualize graph.

Masa929 commented 5 years ago

Thank you for your response and sorry for my late reply. I tried the code below, but an error occurred. Please check it. Thanks.

●Error message File "C:\Users\user\Anaconda3\lib\site-packages\tensorflow\python\framework\graph_util_impl.py", line 137, in _assert_nodes_are_present assert d in name_to_node, "%s is not in graph" % d AssertionError: model/ConvPred/Conv2D:0 is not in graph

●Code python -m tensorflow.python.tools.freeze_graph --input_meta_graph=ckpt/1.ckpt.meta --input_checkpoint=ckpt/1.ckpt --output_graph=ckpt/fcrn.pb --output_node_names="model/ConvPred/Conv2D:0" --input_binary=true

Masa929 commented 5 years ago

I changed "model/ConvPred/Conv2D:0" to "model/ConvPred/Conv2D",Pb file conversion succeeded. Second question, I'd like to know the input node because of quantization. Could you give me the input node? Thanks&Regards.