Train and test on different size datasets

Open saumitrabg opened this issue 5 years ago • 1 comments

We are looking at YOLACT++ inference of 4096x756 image at a high FPS. However, to not run out of memory, our train/val dataset is 756x756. We are seeing low scores when we are running inference on 4096x756. I see the issue is still open: https://github.com/dbolya/yolact/issues/319

Do you have any guidance here?

Dec 20 '20 05:12 saumitrabg

Hey @saumitrabg Did you get any interesting finds on this issue ? , I am facing the same problem as well

Aug 18 '24 03:08 PrajwalCogniac