AI Paper Atlas

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

Levenshtein OCR

Cheng Da, Peng Wang, Cong Yao

ECCV 2022 · Vision-language and multimodal › Human and embodied modeling · approach: Language-grounded scene text and structure · 45 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 iterative refinement from visual-prediction initial sequence (3 iterations)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 90.65 -> 92.81Improved
start from empty initial sequence (3 iterations) (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 90.65 -> 92.63Improved
use visual prediction rather than empty initial sequence (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 92.63 -> 92.81Improved
use random-noise corrupted sequence instead of visual prediction as y0 (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 92.81 -> 85.25Hurt
use ground-truth initial sequence (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 92.81 -> 96.91Improved
second refinement iteration with visual-prediction init (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 92.55 -> 92.79Improved
second iteration with empty initial sequence (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 90.64 -> 92.51Improved
third iteration over second (visual-prediction init) (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST)word accuracy: 92.79 -> 92.81No clear effect
add Levenshtein Transformer deletion and insertion losses (ablation)LevOCR (CNN) · average of 6 benchmarks (MJ+ST), backbone studyword accuracy: 90.70 -> 92.81Improved
add LevT refinement with ViT backboneLevOCR (ViT) · average of 6 benchmarks (MJ+ST), backbone studyword accuracy: 89.30 -> 91.87Improved
use LevOCR instead of SRNscene text recognizer · average of 6 benchmarksword accuracy: 90.4 -> 92.81Improved
use LevOCR instead of ABINetscene text recognizer · average of 6 benchmarksword accuracy: 92.6 -> 92.81Mixed

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