Hello!
Thanks for your great work!
I am wondering how the ablation study is done on the LGAG module, like how does the model fuse the features from skip connections and the output of EUCB. Is it done by simply removing the LGAG, and fuse the skip and EUCB features by simply adding them together which seems not very reasonable because it is not a common way how people aggregate decoder feature and skip features (for example, unet aggregates these two kind of features by concatenating them).
Hello!
Thanks for your great work!
I am wondering how the ablation study is done on the LGAG module, like how does the model fuse the features from skip connections and the output of EUCB. Is it done by simply removing the LGAG, and fuse the skip and EUCB features by simply adding them together which seems not very reasonable because it is not a common way how people aggregate decoder feature and skip features (for example, unet aggregates these two kind of features by concatenating them).