1 paper · 1 filter
Linger Deng, Mingxin Huang, Xudong Xie +3
The advancement of text shape representations towards compactness has enhanced text detection and spotting performance, but at a high annotation cost. Current models use single-poi…