DoPTA pre-trains a vision encoder by aligning image patches with OCR text embeddings using IoU-guided soft labels, improving document layout analysis without OCR at inference.
Gritsenko, Matthias Minderer, Charles Blundell, Razvan Pascanu, and Jovana Mitrovi´c
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
DoPTA: Improving Document Layout Analysis using Patch-Text Alignment
DoPTA pre-trains a vision encoder by aligning image patches with OCR text embeddings using IoU-guided soft labels, improving document layout analysis without OCR at inference.