AI Paper Atlas

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

What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis

Jeonghun Baek, Geewook Kim, Junyeop Lee, Sungrae Park, Dongyoon Han, Sangdoo Yun, Seong Joon Oh, Hwalsuk Lee

ICCV 2019 · Vision-language and multimodal › Human and embodied modeling · approach: Language-grounded scene text and structure · 666 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 TPS transformation (ablation)STR framework (mean over combinations) · Regular benchmarksaccuracy (%): 85.6 -> 86.7Improved
add TPS transformation (ablation)STR framework (mean over combinations) · Irregular benchmarksaccuracy (%): 65.7 -> 69.1Improved
replace VGG with ResNet feature extractor (ablation)STR framework (mean over combinations) · Regular benchmarksaccuracy (%): 84.5 -> 88.3Improved
replace VGG with ResNet feature extractor (ablation)STR framework (mean over combinations) · Irregular benchmarksaccuracy (%): 63.9 -> 71.0Improved
replace VGG with RCNN feature extractor (ablation)STR framework (mean over combinations) · Regular benchmarksaccuracy (%): 84.5 -> 86.2Improved
replace VGG with RCNN feature extractor (ablation)STR framework (mean over combinations) · Irregular benchmarksaccuracy (%): 63.9 -> 67.3Improved
add BiLSTM sequence modeling (ablation)STR framework (mean over combinations) · Regular benchmarksaccuracy (%): 85.1 -> 87.6Improved
add BiLSTM sequence modeling (ablation)STR framework (mean over combinations) · Irregular benchmarksaccuracy (%): 65.2 -> 69.7Improved
replace CTC with attention prediction (ablation)STR framework (mean over combinations) · Regular benchmarksaccuracy (%): 85.5 -> 87.2Improved
replace CTC with attention prediction (ablation)STR framework (mean over combinations) · Irregular benchmarksaccuracy (%): 66.1 -> 68.7Improved
train on MJSynth plus SynthText instead of MJSynth only (ablation)Our best model · Unified benchmark (8,539 images)total accuracy (%): 80.0 -> 84.1Improved
train on MJSynth plus SynthText instead of SynthText only (ablation)Our best model · Unified benchmark (8,539 images)total accuracy (%): 75.6 -> 84.1Improved

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