Abstract: The performance of existing deep-learning scene text recognition-based methods fails significantly on occluded text instances or even partially occluded characters in a text due to their ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results