AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Efficient License Plate Recognition via Holistic Position Attention

Yesheng Zhang, Zilei Wang, Jiafan Zhuang

AAAI 2021 · Architectures and efficiency › Attention and transformers · approach: Efficient image transformers · 14 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
Explicitly encode character positions with HPA maps (ablation)LPR network with ResNet-101 and multiple classifiers · AOLPrecognition accuracy: 83.55 -> 97.65 (AC); 61.17 -> 94.33 (LE); 96.40 -> 99.51 (RP)Improved
Use a shared classifier instead of multiple classifiers (ablation)HPA-based LPR network · AOLPrecognition accuracy: 97.65 -> 99.56 (AC); 94.33 -> 100.00 (LE); 99.51 -> 99.84 (RP)Improved
Use a shared classifier instead of multiple classifiers (ablation)HPA-based LPR network · AOLPmodel size: 71.21 -> 66.77Improved
Train the position network before the semantic network in separate stages (ablation)HPA-based LPR network · AOLPrecognition accuracy: 99.27 -> 90.31 (AC); 99.31 -> 88.49 (LE); 99.51 -> 98.20 (RP)Hurt
Train all networks jointly end to end rather than training them sequentially (ablation)HPA-based LPR network · AOLPrecognition accuracy: 99.27 -> 99.56 (AC); 99.31 -> 100.00 (LE); 99.51 -> 99.84 (RP)Improved
Remove semantic and position supervision signals (ablation)HPA-based LPR network · AOLPrecognition accuracy: 99.56 -> 99.41 (AC); 100.00 -> 99.31 (LE); 99.84 -> 99.67 (RP)Hurt
Use the proposed HPA method with ResNet-101 instead of the strongest reported baselineLicense plate recognition · CCPD-baserecognition accuracy: 99.60 -> 99.65Improved
Use the proposed HPA method with ResNet-101 instead of the strongest reported baselineLicense plate recognition · CLPDrecognition accuracy: 70.80 -> 82.50Improved

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