論文5分鐘 | supervised contrastive learning

摘要: Cross entropy is the most widely used loss function for supervised training of image classification models. In this paper, we propose a novel training methodology that consistently outperforms cro
相關文章
相關標籤/搜索