Multi-Domain Character Distance Perception for Attention-Based Text Recognition
CDistNet: Perceiving Multi-Domain Character Distance for Robust Text Recognition
Attention-based encoder-decoder framework is becoming popular in scenetext recognition, largely due to its superiority in integrating recognition clues from both visual and semantic domains.However, recent studies show the two clues might be misaligned in the difficult text (e.g., with rare textshapes) and introduce constraints such as character position to alleviate the problem.In this paper, we propose a novel module called multi-domain character distance perception(mdcdp) to establish a visual and semantic related position encoding following the attention mechanism.It naturally encodes the positional clue, which describes both visual and semantic distances among characters.We develop a novel architecture named multi-domain character distance perception (cdistnet) that stacks several times to guide precise distance modeling.The experiments demonstrate that the visual-semantic alignment is well built even various difficulties presented.
