Memory Computation and Input Settings in HRDecoder

[ChubbyPear]

Hi, Thank you for your excellent work on the HRDecoder!

I have a question regarding the memory computation: is it calculated specifically when the batch size is set to 1? Is it measured only in the test phase? Additionally, since the DDR and IDRID datasets have different input resolutions and crop numbers, which is used as the standard for the memory computation?

Plus, HRDecoder actually handles a 960x960 downsampled input when hr_scale is set to (960, 960). Which input size did you set for other models when making the comparison in Figure 1a and Table 2?

Thank you for your help!