Is GT_size 160 (size of HR image) hard-coded in the code

[sarvghotra]

1. Is GT_size (size of HR image) hard-coded in the following place:

• https://github.com/andreas128/SRFlow/blob/master/code/models/modules/SRFlowNet_arch.py#L48
• https://github.com/andreas128/SRFlow/blob/master/code/models/modules/FlowUpsamplerNet.py#L280
• https://github.com/andreas128/SRFlow/blob/master/code/models/modules/FlowUpsamplerNet.py#L230
• https://github.com/andreas128/SRFlow/blob/master/code/models/modules/FlowUpsamplerNet.py#L114

2. If yes, does it mean that one can not use any other HR image size during training?
3. Does SRFlow impose any limit on what this hardcarded value could be? In other words, can I use for example 180?

[ghost]

• https://github.com/andreas128/SRFlow/blob/master/code/models/modules/FlowUpsamplerNet.py#L307 is also hard-coded.

I carried out the training after rewriting all these hard-coded values directly.

It would be a good idea to put the image size in the yml files in the directory ./code/confs and change it to refer to that value in all the places in the source code you and I have presented.

If the repository administrator continues to react, I would submit a pull request, but unfortunately no his or her response to these issues seems to be expected.