EISNet
EISNet copied to clipboard
Ablation study (table 4)
Hi, Thank you for your nice work. I am just wondering how can we produce the results deep all, deep all + triplet, deep all + jigsaw, and deep all + triplet + jigsaw. Thank you in advance.
The deepall results are inheriated from paper: Domain Generalization viaModel-Agnostic Learning of Semantic Features. We remove the specific components to generate deep all + triplet, deep all + jigsaw results. And the deep all + triplet + jigsaw is same as our method.