Closed miladm closed 5 years ago
Yes this is correct. There isn't much point on training with higher resolution images unless you have a dataset with more detail. We show in the paper that this resolution is enough to capture the "details" available in the dataset we trained on.
Aaron
MiladM writes:
Hi,
Reading your code, it seems the output resolution is set to support a 192x192 image resolution which implied running it on a higher resolution image requires a new trained model. Can you please confirm? Thanks.
-- Aaron Jackson - M6PIU http://aaronsplace.co.uk/
Hi,
Reading your code, it seems the output resolution is set to support a 192x192 image resolution which implies 3D reconstruction on a higher resolution image requires a new trained model. Can you please confirm? Would love to know if there is a way around it. Thanks.