search

oandrienko / fast-semantic-segmentation

ICNet and PSPNet-50 in Tensorflow for real-time semantic segmentation
220 stars 41 forks source link

For c++ pb #23

Closed engineer1109 closed 5 years ago

engineer1109 commented 5 years ago

Is the input node name "inputs" ? output nodename "Predictions/Conv/Conv2D" ?

And what size and type is the input and output tensor?

oandrienko commented 5 years ago

@engineer1109 The input and output node names for the frozen graph def are set in the export script. I named the input placeholder inputs and the output identity op outputs. The sizes are specified by the export script flags and your model definition. The types are uint for the input and float for the output.

I can make sure to add this to the documentation later and make the input types flags. Hope this helps.

engineer1109 commented 5 years ago

The output node name is predictions_1. the input shape is 1024 2048 3 the output shape is 1024 2048. I have finished my c++ ICNet code . Thank you.

oandrienko commented 5 years ago

Right, the name is predictions for the output - my mistake. Glad it worked!

sctrueew commented 4 years ago

The output node name is predictions_1. the input shape is 1024 2048 3 the output shape is 1024 2048. I have finished my c++ ICNet code . Thank you.

Hi @engineer1109

Could you please share your C++ code?

engineer1109 commented 4 years ago

@zpmmehrdad It has been one year. I need to find my code.

sctrueew commented 4 years ago

@engineer1109 thanks for your reply