AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

OCR-Free Document Understanding Transformer

Geewook Kim, Teakgyu Hong, Moonbin Yim, JeongYeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park

ECCV 2022 · Vision-language and multimodal › Human and embodied modeling · approach: Layout-aware document understanding · 734 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
remove OCR dependence: use end-to-end Donut instead of OCR-based LayoutLMv2Donut · RVL-CDIPaccuracy (%) and time (ms): 95.25 -> 95.30; 1489 -> 752 msImproved
use Donut instead of LayoutLMv2 (OCR-based) for information extractionDonut · CORDfield-level F1 (%): 78.9 -> 84.1Improved
use Donut instead of LayoutLMv2 for information extractionDonut · Tickettree-edit-distance accuracy (%): 90.1 -> 98.7Improved
use end-to-end Donut instead of OCR-based LayoutLMv2Donut · DocVQA testANLS: 78.1 -> 67.5Mixed
use Donut instead of LayoutLMv2-Large-QG on handwritten DocVQA documentsDonut · DocVQA test (handwritten)ANLS: 67.3 -> 72.1Improved
use text reading as pre-training task instead of other tasks such as image captioning (ablation)Donut pre-training · CORD and DocVQAfine-tuning performanceImproved
add real IIT-CDIP images to synthetic SynthDoG pre-training data (ablation)Donut pre-training · DocVQAANLSImproved
add real IIT-CDIP images to synthetic SynthDoG pre-training data (ablation)Donut pre-training · CORDaccuracyNo clear effect
use Swin Transformer or EfficientNetV2 encoder instead of other ImageNet backbones (ablation)Donut encoder · CORD and DocVQAperformanceImproved
increase input resolution (ablation)Donut · CORD and DocVQAaccuracyImproved
use input resolution 1280x960 on CORD (ablation)Donut · CORDspeed and accuracy: 0.7 sec./image and 91.1 accuracyMixed
use larger input resolution 2560x1920 in extremely low-resource training (ablation)Donut · CORD (80 training samples)accuracyImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy