AI Paper Atlas

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

PreSTU: Pre-Training for Scene-Text Understanding

Jihyung Kil, Soravit Changpinyo, Xi Chen, Hexiang Hu, Sebastian Goodman, Wei-Lun Chao, Radu Soricut

ICCV 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Language-grounded scene text and structure · 33 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
add SplitOCR pre-training on CC15Mbaseline ViT-B/16 + mT5-base · TextVQA test-stdaccuracy: 44.8 -> 55.2Improved
add SplitOCR pre-training on CC15Mbaseline · ST-VQA testANLS: 56.7 -> 65.5Improved
add SplitOCR pre-trainingbaseline · VQAv2 test-devaccuracy: 74.8 -> 76.0Improved
add VQA second-stage pre-training after SplitOCRPreSTU-SplitOCR · TextVQA test-stdaccuracy: 55.2 -> 56.3Improved
add SplitOCR then CAP pre-trainingPreSTU-SplitOCR · TextCaps test-stdCIDEr: 126.1 -> 139.1Improved
add SplitOCR pre-trainingbaseline · TextCaps test-stdCIDEr: 96.9 -> 126.1Improved
replace pure OCR pre-training with SplitOCR (stage 1) (ablation)PreSTU · TextVQA valaccuracy: 49.9 -> 55.6Improved
add stage-1 SplitOCR pre-training, evaluated without OCR input at fine-tuning (ablation)PreSTU · TextVQA valaccuracy: 13.7 -> 44.3Improved
add off-the-shelf OCR input at fine-tuning on top of SplitOCR+VQA+CAP (ablation)PreSTU · TextVQA valaccuracy: 44.3 -> 56.7Improved
use 640 instead of 224 pre-training resolution (ablation)PreSTU-SplitOCR · TextVQA valaccuracy: 47.1 -> 55.6Improved
increase pre-training data from 1% to 100% of CC15M (ablation)PreSTU-SplitOCR · TextVQA valaccuracy: 42.3 -> 55.6Improved
use SplitOCR instead of CAP-only second-stage (ablation)PreSTU · TextVQA valaccuracy: 49.3 -> 53.2Improved

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