I want to make sure that in your demo part, gt_y is the ground truth image in YCbCr type and you only choose the y channel; sr_ycbcr is got by the low-resolution image using functional.interpolate(), which is also in YCbCr type; lr_y is the low-resolution image in YCbCr type and you only choose the y channel.
In demo_ATO/save_results(), you replace the y cahnnel of sr_ycbcr with the predicted result.
I want to make sure that in your demo part, gt_y is the ground truth image in YCbCr type and you only choose the y channel; sr_ycbcr is got by the low-resolution image using functional.interpolate(), which is also in YCbCr type; lr_y is the low-resolution image in YCbCr type and you only choose the y channel.
In demo_ATO/save_results(), you replace the y cahnnel of sr_ycbcr with the predicted result.
Is my understanding right? Thank you!