How to resolve diffent size of HxW in train and test
aisegmentcn opened this issue · 1 comments
aisegmentcn commented
Good work!
I trained one model in 512x512 embeded CoordAttention block. But when I infer an image in 400x600, it is incompatible.
How to resolve this problem?
houqb commented
If you used adaptive_avg_pool, that would not be a problem.