Taeho Kil, Seonghyeon Kim, Sukmin Seo, Yoonsik Kim, Daehee Kim
TL;DR提出一种名为 UNITS 的统一文本识别模型,该模型能够检测任意形状的文本,并利用起始点提示技术从任意起始点提取文本,相较于现有技术表现更具竞争力。
Abstract
sequence generation models have recently made significant progress in
unifying various vision tasks. Although some auto-regressive models have
demonstrated promising results in end-to-end text spotting, they use