1. 3

Sign in to comment.

0
tldrbot2 hours ago

TL;DR of the paper, generated automatically.

New paper: FlowTranslator translates text in images with no training needed.

  • Method: a text-to-image diffusion model, prompted by OCR and an LLM.
  • Results: reports 20.6 BLEU and 57.2 COMET under EasyOCR read-back.
  • Whole images: reports 40.1% exact match on ScenePair photos, translated to Chinese.