Toward Advancing License Plate Super-Resolution
in Real-World Scenarios: A Dataset and Benchmark

Recent advancements in super-resolution for License Plate Recognition (LPR) have sought to address challenges posed by low-resolution (LR) and degraded images in surveillance, traffic monitoring, and forensic applications. However, existing studies have relied on private datasets and simplistic degradation models. To address this gap, we introduce UFPR-SR-Plates, a novel dataset containing 10,000 tracks with 100,000 paired low and high-resolution license plate images captured under real-world conditions. We establish a benchmark using multiple sequential LR and high-resolution (HR) images per vehicle – five of each – and two state-of-the-art models for super-resolution of license plates. We also investigate three fusion strategies to evaluate how combining predictions from a leading Optical Character Recognition (OCR) model for multiple super-resolved license plates enhances overall performance. Our findings demonstrate that super-resolution significantly boosts LPR performance, with further improvements observed when applying majority vote-based fusion techniques. Specifically, the Layout-Aware and Character-Driven Network (LCDNet) model combined with the Majority Vote by Character Position (MVCP) strategy led to the highest recognition rates, increasing from 1.7% with low-resolution images to 31.1% with super-resolution, and up to 44.7% when combining OCR outputs from five super-resolved images. These findings underscore the critical role of super-resolution and temporal information in enhancing LPR accuracy under real-world, adverse conditions. The proposed dataset is publicly available to support further research.

Authors: Valfride Nascimento, Gabriel E. Lima, Rafael O. Ribeiro, William Robson Schwartz, Rayson Laroca, David Menotti

Title: Toward Advancing License Plate Super-Resolution in Real-World Scenarios: A Dataset and Benchmark

Summary:

In summary, the main contributions of this work are:

• A publicly available dataset containing 100,000 images, divided into 10,000 tracks, with each track containing five LR images and five HR images of the same LP. To enhance variability, 5,000 tracks were collected at a resolution of 1280 × 960 pixels, while the remaining 5,000 tracks were captured at 1920 × 1080 pixels. The dataset is evenly distributed between Mercosur and Brazilian LPs, making it the largest dataset in terms of the number of LPs for both layouts;
• We conducted benchmark experiments on the proposed dataset using five state-of-the-art super-resolution models: (i) general-purpose approaches (SR3 [Sahariaet al., 2023], Real-ESRGAN [Wang et al., 2021]), and (ii) LP-specialized networks (LPSRGAN [Pan et al., 2024], PLNET [Nascimento et al., 2024a], and LCDNet [Nascimento et al., 2024b]). For each track, we generated five super-resolved images from the LR images and compared the recognition results obtained by the leading Optical Character Recognition (OCR) model, GP_LPR [Liu et al., 2024b]. The super-resolution process significantly boosted recognition accuracy, increasing from 2.2% to 29.9% for a single super-resolved image. To further enhance LPR performance, we explored three fusion strategies for combining the outputs from the OCR model based on multiple super-resolved images. Notably, applying the Majority Vote by Character Position (MVCP) strategy with five super-resolved images improved the recognition rate from 29.9% to 42.3%. The proposed dataset enables the exploration of temporal relationships among low-resolution LPs, as it includes multiple sequential LR images for each LP.

This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001, and in part by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)(# 315409/2023-1 and # 312565/2023-2). We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Quadro RTX 8000 GPU used for this research.