Real-300K
Real-300K copied to clipboard
The dataset used in the CVPR 2022 paper (SimAN: Exploring Self-Supervised Representation Learning of Scene Text via Similarity-Aware Normalization).
Results
2
Real-300K issues
Sort by
recently updated
recently updated
newest added
你好,最近在看这篇论文SimAN,有一个问题一直很疑惑,论文中两个特征(来自 I_s 和 I_aug), 通过 instance norm 产生了 key和 query; 二者来自图像的不同部分,都属于content ; 但已疑惑的是 key和query做 矩阵乘法的意义是什么,有什么可以解释的通的地方?
Thank you for the wonderful work. Any plan to release the code for SimAN pre-training?