LPLCv2: An Expanded Dataset for Fine-Grained License Plate Legibility Classification

Lucas Wojcik; Eduardo A. F. Machoski; Eduil Nascimento; Rayson Laroca; David Menotti

cs.CVApr 9, 2026

LPLCv2: An Expanded Dataset for Fine-Grained License Plate Legibility Classification

Authors: Lucas Wojcik, Eduardo A. F. Machoski, Eduil Nascimento, Rayson Laroca, David Menotti

Organizations: Federal University of Paran´a, Curitiba, Brazil · Department of Technological Development and Quality, Paran´a Military Police, Curitiba, Brazil · Pontifical Catholic University of Paran´a, Curitiba, Brazil

Abstract

Modern Automatic License Plate Recognition (ALPR) systems achieve outstanding performance in controlled, well-defined scenarios. However, large-scale real-world usage remains challenging due to low-quality imaging devices, compression artifacts, and suboptimal camera installation. Identifying illegible license plates (LPs) has recently become feasible through a dedicated benchmark; however, its impact has been limited by its small size and annotation errors. In this work, we expand the original benchmark to over three times the size with two extra capture days, revise its annotations and introduce novel labels. LP-level annotations include bounding boxes, text, and legibility level, while vehicle-level annotations comprise make, model, type, and color. Image-level annotations feature camera identity, capture conditions (e.g., rain and faulty cameras), acquisition time, and day ID. We present a novel training procedure featuring an Exponential Moving Average-based loss function and a refined learning rate scheduler, addressing common mistakes in testing. These improvements enable a baseline model to achieve an 89.5% F1-score on the test set, considerably surpassing the previous state of the art. We further introduce a novel protocol to explicitly addresses camera contamination between training and evaluation splits, where results show a small impact. Dataset and code are publicly available at https://github.com/lmlwojcik/LPLCv2-Dataset.

Figures & tables

Panel of Figure 1 — Fig. 1: LPs grouped by legibility class, following the definition of [ 10 ] .

Panel of Figure 2 — Fig. 2: Representative samples from the proposed LPLCv2 dataset.

Images by Time of Day		Images by Attributes
Class	Number	Attribute	Number
Morning	$10{\mathord{\mathchar 59\relax}}998$	Faulty Camera	$\phantom{0}3{\mathord{\mathchar 59\relax}}690$
Afternoon	$12{\mathord{\mathchar 59\relax}}799$	Raining	$\phantom{00\mathord{\mathchar 59\relax}}770$
Evening	$\phantom{0}9{\mathord{\mathchar 59\relax}}157$	Has Camera ID	$29{\mathord{\mathchar 59\relax}}965$
Night	$\phantom{0}4{\mathord{\mathchar 59\relax}}145$	$\rightarrow$ Total Images	$37{\mathord{\mathchar 59\relax}}099$

TABLE I: LPLCv2 Image Statistics.

LPs by Legibility		Other Attributes
Class	Number	Attribute	Number
Perfect	$18{\mathord{\mathchar 59\relax}}425$	Vehicle Visible	$38{\mathord{\mathchar 59\relax}}089$
Good	$10{\mathord{\mathchar 59\relax}}180$	Vehicle Available	$25{\mathord{\mathchar 59\relax}}506$
Poor	$\phantom{0}7{\mathord{\mathchar 59\relax}}520$	LP Text Available	$36{\mathord{\mathchar 59\relax}}414$
Illegible	$\phantom{0}5{\mathord{\mathchar 59\relax}}362$	$\rightarrow$ Total LPs	$41{\mathord{\mathchar 59\relax}}487$

TABLE II: LPLCv2 LP Statistics.

Fig. 3: LP legibility annotation errors observed in LPLCv1 (left) and their corresponding corrections in LPLCv2 (right).

Scenario	Classes
Baseline	Perfect	Good	Poor	Illegible
Legibility Recognition	Suitable		Poor	–
Full Recognition	Suitable		Poor	Illegible
Quality Filter	Suitable		Unsuitable

TABLE III: Mapping of original legibility classes to labels used in each evaluation scenario.

Partition	Class				Overall
Partition	Perfect	Good	Poor	Illegible	Overall
LPLCv1	$84.5$ %	$68.0$ %	$56.7$ %	$73.0$ %	$74.5$ %
+ Revised Labels	$92.1$ %	$83.1$ %	$83.8$ %	$88.2$ %	$84.4$ %
+ More Images	$92.8$ %	$86.0$ %	$86.0$ %	$92.9$ %	$88.7$ %
+ EMA	$92.7$ %	$86.2$ %	$86.3$ %	$93.0$ %	$89.8$ %

TABLE IV: Baseline results using ResNet-50, reported as F1-score (%).

Fig. 4: Confusion matrix from one of the EMA experiment folds.

Fig. 5: Examples of misclassified LPs illustrating borderline legibility cases.

Panel of Figure 10 — Fig. 6: LPs misclassified into non-adjacent legibility classes.

Partition	Legibility	Full	Quality
Partition	Recognition	Recognition	Filter
LPLCv1	$92.6$ %	$87.2$ %	$93.3$ %
+ Revised Labels	$95.3$ %	$93.7$ %	$96.4$ %
+ More Images	$95.9$ %	$95.2$ %	$96.9$ %
+ EMA	$96.5$ %	$94.8$ %	$96.7$ %

TABLE V: ResNet-50 performance (F1-score) in the remaining scenarios: Legibility Recognition (Suitable and Poor); Full Recognition (Suitable, Poor and Illegible); and Quality Filter (Suitable and Unsuitable).

Partition	Class				Overall
Partition	Perfect	Good	Poor	Illegible	Overall
Intra-Camera	$93.4$ %	$82.6$ %	$84.8$ %	$91.6$ %	$89.2$ %
Cross-Camera	$93.5$ %	$81.8$ %	$85.7$ %	$91.1$ %	$88.2$ %

TABLE VI: Baseline scenario results under intra-camera and cross-camera protocols (F1-score).

Partition	Legibility	Full	Quality
Partition	Recognition	Recognition	Filter
Intra-Camera	$96.2$ %	$94.5$ %	$96.6$ %
Cross-Camera	$95.6$ %	$93.8$ %	$96.2$ %

TABLE VII: ResNet-50 performance across the other evaluation scenarios under intra-camera and cross-camera splits (F1-score).

Automatic License Plate Recognition

Automatic license plate recognition detects vehicle plates and reads their characters from images or video for traffic and access-control systems.

See all 21 papers

ICPR 2026 Competition on Low-Resolution License Plate Recognition
Apr 24, 2026Rayson Laroca, Valfride Nascimento, Donggun Kim +19Automatic License Plate Recognition
Federal University of Paraná, Brazil · Pontifical Catholic University of Paraná, Brazil · Korea University, Republic of Korea +6
Street-Legal Physical-World Adversarial Rim for License Plates
Apr 2, 2026Nikhil Kalidasu, Sahana GanapathyAdversarial Attacks Physical Adversarial Attacks
The University of Texas at Austin, Austin TX 78705, USA
Toward Unified Fine-Grained Vehicle Classification and Automatic License Plate Recognition
Apr 7, 2026Gabriel E. Lima, Valfride Nascimento, Eduardo Santos +3Fine-Grained Image Classification Fine-Grained Visual Perception
Department of Informatics, Federal University of Paraná, R. Evaristo F. Ferreira da Costa 391, Jardim das Américas, Curitiba, PR, 81530-090, Brazil.

Apr 24, 2026J/K move · Enter open · S save

ICPR 2026 Competition on Low-Resolution License Plate Recognition

Rayson Laroca, Valfride Nascimento, Donggun Kim, Sanghyeok Chung+18

Federal University of Paraná, Brazil · Pontifical Catholic University of Paraná, Brazil · Korea University, Republic of Korea+6

Automatic License Plate Recognition

Open paper AskPDF arXiv

Low-Resolution License Plate Recognition (LRLPR) remains a challenging problem in real-world surveillance scenarios, where long capture distances, compression artifacts, and adverse imaging conditions can severely degrade license plate legibility. To promote progress in this area, we organized the ICPR 2026 Competition on Low-Resolution License Plate Recognition, the first competition specifically dedicated to LRLPR using real low-quality data collected under operationally relevant conditions. The competition was based on the LRLPR-26 dataset, which comprises 20,000 training tracks and 3,000 test tracks; each training track contains five low-resolution and five high-resolution images of the same license plate. Notably, a total of 269 teams from 41 countries registered for the competition, and 99 teams submitted valid entries in the Blind Test Phase. The winning team achieved a Recognition Rate of 82.13%, and four teams surpassed the 80% mark, highlighting both the high level of competition at the top of the leaderboard and the continued difficulty of the task. In addition to presenting the competition design, evaluation protocol, and main results, this paper summarizes the methods adopted by the top-5 teams and discusses current trends and promising directions for future research on LRLPR. The competition webpage is available at https://icpr26lrlpr.github.io/

Abstract

Automatic License Plate Recognition

Explore similar work

ICPR 2026 Competition on Low-Resolution License Plate Recognition

Street-Legal Physical-World Adversarial Rim for License Plates

Toward Unified Fine-Grained Vehicle Classification and Automatic License Plate Recognition