The annotation-budget study shows that a few hundred labeled localization annotations substantially improve the conventional pipeline, enabling it to surpass zero-shot VLMs in character recovery, although VLMs retain an advantage in image-level plate counting.
Abstract
Automatic license-plate recognition (ALPR) systems typically rely on a two-stage pipeline that combines object detection with optical character recognition (OCR), whereas recent vision–language models (VLMs) enable direct image-level recognition without task-specific training. This paper presents a comprehensive comparison between these paradigms under limited target-domain data availability. We evaluate a plate-specific YOLO11x detector combined with four OCR backends and five locally hosted open-weight VLMs, without training or fine-tuning on a study-specific Serbian dataset containing 435 images and 643 visible license plates, annotated with image-level plate counts, registration transcriptions, and plate-level bounding boxes. Performance is assessed using strict end-to-end success, exact-match accuracy, character error rate, readable-plate recall, plate-count accuracy, and inference speed. The comparison is further validated on the OpenALPR-EU and UC3M-LP benchmarks. Finally, we investigate the effect of increasing target-domain annotation budgets by incrementally fine-tuning the YOLO detector on UC3M-LP to determine when a conventional detector–OCR pipeline overtakes zero-shot VLMs. The results demonstrate complementary strengths rather than a universal winner. YOLO11x combined with FastPlateOCR achieves the highest transcription accuracy and is approximately 13× faster than the best-performing VLM; the three general-purpose OCR backends collapse on identical detections, so these figures characterize the strongest available plate-specific recognizer rather than detector–OCR pipelines in general, while Qwen3-VL provides superior image-level plate-count accuracy in the zero-shot setting, an advantage concentrated in single-plate scenes. The annotation-budget study shows that a few hundred labeled localization annotations substantially improve the conventional pipeline, enabling it to surpass zero-shot VLMs in character recovery, although VLMs retain an advantage in image-level plate counting. These findings provide practical guidance for selecting ALPR systems according to annotation availability, deployment constraints, and application objectives.
Large language models (LLMs) with multimodal input capability have been demonstrated in various applications. Vision language models (VLMs), which were developed from LLMs, can process and learn both visual and textual data simultaneously, enabling them to generate descriptive text from images. This study explores the...
A. V. Ega, Dadang Gunawan, Abdul Halim et al.· Bulletin of Electrical Engin...· 0 citations
This study presents an image-quality-aware framework that integrates YOLOv11n plate localization, selective CLAHE enhancement, PP-OCRv5 recognition, and syntax-constrained post-processing and identifies syntax-constrained correction as the principal contributor.
Shao-Rong Li, Zhi-Qin Yu, Lan Xu et al.· Applied Sciences· 0 citations
Detecting, localizing, and reading a license plate in a single pass usually lowers accuracy. Each of these sub-tasks is manageable on its own, but a single model must handle all of them together, across spatial scales that differ by more than an order of magnitude, so it settles for a compromise that is only moderate a...
Radha Krishna Gajurel, Agrim Sigdel, Bijaya Thapa Magar et al.· National College of Computer...· 0 citations
VDiff-Bench provides a targeted diagnostic for evaluating comparative visual understanding in MLLMs, exposing failures that are not captured by standard single-image vision-language tasks.
Comparisons show that OCR specialization largely preserves head identity, retaining 17-20 of the top 20 heads per task with all-head rank correlations of 0.874-0.942, while redistributing their functional and causal strengths.
Yuan-Xiang Huangfu, Han-Meng Zhong, Linqing Chen et al.· 0 citations
An approximately 170M-sample OCR-centric corpus is built using an automated data engine that combines expert consensus, render-based verification, and targeted synthesis and shows that, with sufficient parsing training, OCR-centric understanding supervision provides additional gains for document parsing.
Xin Chen, An-An Du, Feng-Juan Feng et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.