Hello, thank you for your great work.
Here, I have a question about model building.
In the file "TrainerNormal.py", the function "_build_normal_refiner" contains the details of normal refinement network.
The encoder is from line 605-614, and the decoder is from line 624-637.
But at line 631, why use "slim.conv2d" instead of "slim.conv2d_transpose" ?
Hello, thank you for your great work.
Here, I have a question about model building.
In the file "TrainerNormal.py", the function "_build_normal_refiner" contains the details of normal refinement network.
The encoder is from line 605-614, and the decoder is from line 624-637.
But at line 631, why use "slim.conv2d" instead of "slim.conv2d_transpose" ?