Input and output sizes should be greater than 0, but got input (H: 16, W: 16) output (H: 0, W: 0)

[icey-zhang]

Input and output sizes should be greater than 0, but got input (H: 16, W: 16) output (H: 0, W: 0)

[icey-zhang]

You can refer to icey-zhang/SuperYOLO#25

[melika-sce]

@icey-zhang Thank you, I follow the solution at that issue and it worked, but I have a question
I want to train SRyolo with both train_image_size and test_image_size equal to 1024, but when I set the down_factor= 2 in code (this was the solution provided in that issue), my train config shows different information

is it possible to train SuperYolo with same train and test size?
and which of these information in the log is correct?
is the image size 512x512 or 1024x1024 or anything else?

Thanks

[icey-zhang]

Please set the train_image_size=1024, test_image_size=512 to use the super resolution branch

[melika-sce]

so the train_image_size must be double of test_image_size, right?
can I set train_image_size=2048 and test_image_size=1024 to use the super resolution branch too?

[icey-zhang]

it is okay, and the code can run any factor. In our experiment, the 2x is the best！

[melika-sce]

Thanks!