Transformer and Deep Learning Applications in Image Processing

Journal Name	Impact Factor	CiteScore	Launched Year	First Decision (median)	APC
Diagnostics diagnostics	3.8	6.9	2011	20.4 Days	CHF 2600
Electronics electronics	2.9	7.0	2012	14.8 Days	CHF 2400
Journal of Imaging jimaging	3.8	7.3	2015	21.3 Days	CHF 1800
Mathematics mathematics	2.3	5.4	2013	17.4 Days	CHF 2600
Sensors sensors	4.0	9.4	2001	17.8 Days	CHF 2600

24 pages, 2439 KB

Open AccessArticle

Forecasting Artificial Intelligence News Sentiment Index: Traditional vs. Image-Based Deep-Learning Models

by Gianina-Maria Petrașcu, Ioana Bîrlan, Cristina-Rodica Boboc and Adriana AnaMaria Davidescu

Mathematics 2026, 14(15), 2789; https://doi.org/10.3390/math14152789 - 4 Aug 2026

Viewed by 436

Rapid AI adoption has intensified public and media attention toward AI-related developments, making news sentiment an increasingly important indicator of expectations and perceptions. This study constructs a global daily AI News Sentiment Index using data from the GDELT Global Knowledge Graph and examines [...] Read more.

Rapid AI adoption has intensified public and media attention toward AI-related developments, making news sentiment an increasingly important indicator of expectations and perceptions. This study constructs a global daily AI News Sentiment Index using data from the GDELT Global Knowledge Graph and examines the forecasting properties of the index using statistical, sequential deep learning and image-based forecasting methods. The period of observation of the dataset spans from January 2016 to December 2025, and the sample consists of 3635 daily observations. A sentiment index is constructed from the positive and negative sentiments and is analyzed together with its constituent series. For the empirical framework, ARIMA, ARIMA–GARCH, ETS, LightGBM, LSTM, CNN, TCN, GAN, and convolutional models based on Gramian Angular Summation Field (GASF) and Gramian Angular Difference Field (GADF) transformations are compared using accuracy evaluation on the out-of-sample test set. In this study, we find no evidence that increasing model complexity leads to better forecasting performance for the AI news sentiment series. In most cases, sequential forecasting models perform better than their image-based counterparts, while GASF and GADF transformations do not deliver consistently better forecasting performance for sentiment series of different types and different forecasting horizons. This result indicates that transforming noisy sentiment time series into an image may hide rather than preserve useful information for forecasting purposes. The study contributes to the growing literature on AI news sentiment forecasting by providing a comprehensive comparison of statistical, sequential, and image-based forecasting paradigms and offers practical insights for researchers, policymakers, and practitioners interested in monitoring AI-related expectations and sentiment dynamics. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

25 pages, 7278 KB

Open AccessArticle

An Adaptive Edge-Guided Dual-Network Framework for Fast QR Code Motion Deblurring

by Jianping Li, Dongyang Guo, Wenjie Li and Wei Zhao

Sensors 2026, 26(15), 4879; https://doi.org/10.3390/s26154879 - 3 Aug 2026

Viewed by 507

Abstract

Unlike natural image deblurring, which primarily emphasizes perceptual quality and pixel-level fidelity, Quick Response (QR) code deblurring must retain decoding-critical structures to guarantee successful decoding. QR codes contain regular binary module grids and functional patterns with sharp boundaries, providing a strong structural prior [...] Read more.

Unlike natural image deblurring, which primarily emphasizes perceptual quality and pixel-level fidelity, Quick Response (QR) code deblurring must retain decoding-critical structures to guarantee successful decoding. QR codes contain regular binary module grids and functional patterns with sharp boundaries, providing a strong structural prior for restoration. However, most existing learning-based QR restoration methods capture QR-specific structural information via implicit feature learning. To address this limitation, we propose an Edge-Guided Attention Block (EGAB), which explicitly extracts multi-directional edge priors and injects them into the query–key correlations of Transformer attention. Based on EGAB, we develop an Edge-Guided Restormer (EG-Restormer) for restoring severely blurred QR codes. For mildly blurred inputs, we introduce a Lightweight and Efficient Network (LENet) that performs fast restoration with low computational overhead. We further integrate EG-Restormer and LENet into an Adaptive Dual-network (ADNet), which selects the appropriate restoration branch according to the input blur level. Extensive experiments demonstrate the effectiveness of the proposed framework. EG-Restormer boosts the decoding rate by 8.67 percentage points under GoPro-only training and achieves the highest decoding rate among the evaluated methods after QRData fine-tuning. Moreover, ADNet reduces average inference latency by 19% while maintaining comparable decoding performance. These results suggest that explicit edge prior modeling enhances the recovery of structures critical for decoding, while adaptive routing provides an effective balance between decoding accuracy and computational efficiency. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

44 pages, 17314 KB

Open AccessSystematic Review

Industrial Object Counting from Traditional Machine Vision to Open-World Foundation Models: A Systematic Review

by Wei Wang, Shengjie Zhang, Jin He, Lanhui Liu, Wu Du and Le Zhang

Sensors 2026, 26(14), 4494; https://doi.org/10.3390/s26144494 - 15 Jul 2026

Viewed by 1075

Abstract

As a fundamental and highly challenging task in the field of computer vision, industrial object counting plays a critical role in smart manufacturing, inventory management, and production process monitoring. Over the past fifteen years (2010–2025), this field has undergone a profound technological transformation, [...] Read more.

As a fundamental and highly challenging task in the field of computer vision, industrial object counting plays a critical role in smart manufacturing, inventory management, and production process monitoring. Over the past fifteen years (2010–2025), this field has undergone a profound technological transformation, shifting from traditional machine vision methods relying on handcrafted features to a data-driven paradigm based on deep learning. This paper aims to provide a comprehensive and systematic review of this rapidly evolving research area, with technological evolution as the core narrative thread. First, we review early traditional methods, analyzing the application of sensor-based and template-matching technologies in controlled environments, as well as their core limitations in complex industrial scenarios. Subsequently, this paper focuses on exploring how the introduction of deep learning has reshaped the landscape of counting tasks, and elaborates on the breakthrough progress of convolutional neural networks (CNNs), Transformer architectures, the recently emerging Mamba state space model, and Large Foundation Models in addressing key challenges including occlusion, object overlap, multi-scale variation, and dense object counting. In particular, this paper conducts an in-depth analysis of the paradigm shift from Class-Specific Counting to Class-Agnostic Counting (CAC) and Exemplar-Free Counting. This trend significantly reduces the reliance on large-scale annotated data and greatly enhances the generalization ability of models in open-world scenarios. Additionally, this paper systematically organizes mainstream datasets in the field, including FSC-147, NWPU-MOC, and OmniCount-191, and compares core evaluation metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the PrACo metric system. In response to the core technical challenges faced by current methods, including high annotation costs, weak cross-domain adaptability, and strict real-time requirements in industrial scenarios, this paper proposes key future research directions including lightweight model design, unsupervised learning, multi-modal fusion, and Prompt-based interactive counting. This review intends to provide researchers in both academia and industry with a complete technical blueprint so as to promote the continuous development of industrial object-counting technology toward a more efficient and intelligent direction. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

18 pages, 1889 KB

Open AccessArticle

Vision Transformer with Spatial 2D Multi-Channel Tokens

by Sirui Zheng, Yu Li, Zhongxiang Zhang and Dequn Zhao

Electronics 2026, 15(13), 2752; https://doi.org/10.3390/electronics15132752 - 23 Jun 2026

Viewed by 465

Abstract

Vision Transformer (ViT) has been widely adopted in the computer vision community. However, the standard ViT often contains many parameters, usually performs poorly when trained from scratch on medium-scale datasets, and does not explicitly preserve the local spatial and channel-wise structures within each [...] Read more.

Vision Transformer (ViT) has been widely adopted in the computer vision community. However, the standard ViT often contains many parameters, usually performs poorly when trained from scratch on medium-scale datasets, and does not explicitly preserve the local spatial and channel-wise structures within each token. This work proposes a novel model called the Token-Shared Convolutional Projection Vision Transformer (TSCP-ViT). The core idea of TSCP-ViT is to integrate convolutional layers into the multi-head attention mechanism and to apply the same convolutional operation independently to each token, where each token exhibits spatial 2D multi-channel characteristics. In addition, this work introduces a Transformer decoder immediately after each Transformer encoder, enabling the classification tokens to aggregate information from all tokens and be updated using statistical information. Moreover, a trainable Non-Reversing Gate GELU (NRG-GELU) activation is also proposed. Comparative experiments on CIFAR-100, Food-101, and ImageNet100 show that, under comparable parameter counts and without pretraining or knowledge distillation, TSCP-ViT substantially surpasses ViT, outperforms CvT, outperforms ResNet on Food-101, and approaches ResNet on CIFAR-100 and ImageNet100, although with considerably higher FLOPs. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

18 pages, 6162 KB

Open AccessArticle

YOLO-UTD: A Domain-Specific Detection Framework for Small Objects in UAV Traffic Surveillance

by Hailang Huang, Meng Li, Jiebao Zhang and Yitong Li

Sensors 2026, 26(12), 3931; https://doi.org/10.3390/s26123931 - 20 Jun 2026

Viewed by 673

Abstract

Detecting objects in drone-captured aerial imagery is particularly formidable due to challenges such as the prevalence of numerous small targets and their dense spatial distribution. To bridge this gap, this paper introduces YOLO-UTD (YOLO-UAV Traffic Detection), a dedicated small object detector tailored for [...] Read more.

Detecting objects in drone-captured aerial imagery is particularly formidable due to challenges such as the prevalence of numerous small targets and their dense spatial distribution. To bridge this gap, this paper introduces YOLO-UTD (YOLO-UAV Traffic Detection), a dedicated small object detector tailored for drone traffic surveillance. Built upon the YOLOv8 framework, the proposed model incorporates three principal enhancements. First, a specialized small-object detection head replaces the original large-object head to increase the sensitivity to fine-grained features. Second, we introduce a shallow-augmented feature pyramid network (SFPN) into the neck module. The SFPN enriches the semantic content of high-resolution shallow features via dense multiscale interactions and CARAFE upsampling, boosting performance on small targets. Finally, a C2fA layer is integrated into the deep backbone stages to adaptively fuse spatial details and semantic context through a dual-path architecture and a cross-attention mechanism, thereby dynamically refining features critical for small objects. Extensive experiments on the VisDrone2019 dataset validate that YOLO-UTD achieves a 3.6% higher mean average precision (mAP) than YOLOv8 while preserving a low parameter footprint, with a particularly significant gain of 5.3% in vehicle detection accuracy. These findings confirm the model’s efficacy and strong potential for application in smart city drone surveillance. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

24 pages, 3126 KB

Open AccessArticle

Calibrated Transformer Fusion for Dual-View Low-Energy CESM Classification

by Ahmed A. H. Alkurdi and Amira Bibo Sallow

J. Imaging 2026, 12(1), 41; https://doi.org/10.3390/jimaging12010041 - 13 Jan 2026

Cited by 2 | Viewed by 1099

Abstract

Contrast-enhanced spectral mammography (CESM) provides low-energy images acquired in standard craniocaudal (CC) and mediolateral oblique (MLO) views, and clinical interpretation relies on integrating both views. This study proposes a dual-view classification framework that combines deep CNN feature extraction with transformer-based fusion for breast-side [...] Read more.

Contrast-enhanced spectral mammography (CESM) provides low-energy images acquired in standard craniocaudal (CC) and mediolateral oblique (MLO) views, and clinical interpretation relies on integrating both views. This study proposes a dual-view classification framework that combines deep CNN feature extraction with transformer-based fusion for breast-side classification using low-energy (DM) images from CESM acquisitions (Normal vs. Tumorous; benign and malignant merged). The evaluation was conducted using 5-fold stratified group cross-validation with patient-level grouping to prevent leakage across folds. The final configuration (Model E) integrates dual-backbone feature extraction, transformer fusion, MC-dropout inference for uncertainty estimation, and post hoc logistic calibration. Across the five held-out test folds, Model E achieved a mean accuracy of 96.88% ± 2.39% and a mean F1-score of 97.68% ± 1.66%. The mean ROC-AUC and PR-AUC were 0.9915 ± 0.0098 and 0.9968 ± 0.0029, respectively. Probability quality was supported by a mean Brier score of 0.0236 ± 0.0145 and a mean expected calibration error (ECE) of 0.0334 ± 0.0171. An ablation study (Models A–E) was also reported to quantify the incremental contribution of dual-view input, transformer fusion, and uncertainty calibration. Within the limits of this retrospective single-center setting, these results suggest that dual-view transformer fusion can provide strong discrimination while also producing calibrated probabilities and uncertainty outputs that are relevant for decision support. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

22 pages, 1777 KB

Open AccessArticle

DP2PNet: Diffusion-Based Point-to-Polygon Conversion for Single-Point Supervised Oriented Object Detection

by Peng Li, Limin Zhang and Tao Qu

Sensors 2026, 26(1), 329; https://doi.org/10.3390/s26010329 - 4 Jan 2026

Viewed by 819

Abstract

Rotated Bounding Boxes (RBBs) for oriented object detection are labor-intensive and time-consuming to annotate. Single-point supervision offers a cost-effective alternative but suffers from insufficient size and orientation information, leading existing methods to rely heavily on complex priors and fixed refinement stages. In this [...] Read more.

Rotated Bounding Boxes (RBBs) for oriented object detection are labor-intensive and time-consuming to annotate. Single-point supervision offers a cost-effective alternative but suffers from insufficient size and orientation information, leading existing methods to rely heavily on complex priors and fixed refinement stages. In this paper, we propose DP2PNet (Diffusion-Point-to-Polygon Network), the first diffusion model-based framework for single-point supervised oriented object detection. DP2PNet features three key innovations: (1) A multi-scale consistent noise generator that replaces manual or external model priors with Gaussian noise, reducing dependency on domain-specific information; (2) A Noise Cross-Constraint module based on multi-instance learning, which selects optimal noise point bags by fusing receptive field matching and object coverage; (3) A Semantic Key Point Aggregator that aggregates noise points via graph convolution to form semantic key points, from which pseudo-RBBs are generated using convex hulls. DP2PNet supports dynamic adjustment of refinement stages without retraining, enabling flexible accuracy optimization. Extensive experiments on DOTA-v1.0 and DIOR-R datasets demonstrate that DP2PNet achieves 53.82% and 53.61% mAP50, respectively, comparable to methods relying on complex priors. It also exhibits strong noise robustness and cross-dataset generalization. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

18 pages, 2632 KB

Open AccessArticle

Adverse-Weather Image Restoration Method Based on VMT-Net

by Zhongmin Liu, Xuewen Yu and Wenjin Hu

J. Imaging 2025, 11(11), 376; https://doi.org/10.3390/jimaging11110376 - 26 Oct 2025

Viewed by 1890

Abstract

To address global semantic loss, local detail blurring, and spatial–semantic conflict during image restoration under adverse weather conditions, we propose an image restoration network that integrates Mamba with Transformer architectures. We first design a Vision-Mamba–Transformer (VMT) module that combines the long-range dependency modeling [...] Read more.

To address global semantic loss, local detail blurring, and spatial–semantic conflict during image restoration under adverse weather conditions, we propose an image restoration network that integrates Mamba with Transformer architectures. We first design a Vision-Mamba–Transformer (VMT) module that combines the long-range dependency modeling of Vision Mamba with the global contextual reasoning of Transformers, facilitating the joint modeling of global structures and local details, thus mitigating information loss and detail blurring during restoration. Second, we introduce an Adaptive Content Guidance (ACG) module that employs dynamic gating and spatial–channel attention to enable effective inter-layer feature fusion, thereby enhancing cross-layer semantic consistency. Finally, we embed the VMT and ACG modules into a U-Net backbone, achieving efficient integration of multi-scale feature modeling and cross-layer fusion, significantly improving reconstruction quality under complex weather conditions. The experimental results show that on Snow100K-S/L, VMT-Net improves PSNR over the baseline by approximately 0.89 dB and 0.36 dB, with SSIM gains of about 0.91% and 0.11%, respectively. On Outdoor-Rain and Raindrop, it performs similarly to the baseline and exhibits superior detail recovery in real-world scenes. Overall, the method demonstrates robustness and strong detail restoration across diverse adverse-weather conditions. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

21 pages, 6219 KB

Open AccessArticle

Model-Free Transformer Framework for 6-DoF Pose Estimation of Textureless Tableware Objects

by Jungwoo Lee, Hyogon Kim, Ji-Wook Kwon, Sung-Jo Yun, Na-Hyun Lee, Young-Ho Choi, Goobong Chung and Jinho Suh

Sensors 2025, 25(19), 6167; https://doi.org/10.3390/s25196167 - 5 Oct 2025

Cited by 2 | Viewed by 1767

Abstract

Tableware objects such as plates, bowls, and cups are usually textureless, uniform in color, and vary widely in shape, making it difficult to apply conventional pose estimation methods that rely on texture cues or object-specific CAD models. These limitations present a significant obstacle [...] Read more.

Tableware objects such as plates, bowls, and cups are usually textureless, uniform in color, and vary widely in shape, making it difficult to apply conventional pose estimation methods that rely on texture cues or object-specific CAD models. These limitations present a significant obstacle to robotic manipulation in restaurant environments, where reliable six-degree-of-freedom (6-DoF) pose estimation is essential for autonomous grasping and collection. To address this problem, we propose a model-free and texture-free 6-DoF pose estimation framework based on a transformer encoder architecture. This method uses only geometry-based features extracted from depth images, including surface vertices and rim normals, which provide strong structural priors. The pipeline begins with object detection and segmentation using a pretrained video foundation model, followed by the generation of uniformly partitioned grids from depth data. For each grid cell, centroid positions, and surface normals are computed and processed by a transformer-based model that jointly predicts object rotation and translation. Experiments with ten types of tableware demonstrate that the method achieves an average rotational error of 3.53 degrees and a translational error of 13.56 mm. Real-world deployment on a mobile robot platform with a manipulator further validated its ability to autonomously recognize and collect tableware, highlighting the practicality of the proposed geometry-driven approach for service robotics. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

20 pages, 2226 KB

Open AccessArticle

RST-Net: A Semantic Segmentation Network for Remote Sensing Images Based on a Dual-Branch Encoder Structure

by Na Yang, Chuanzhao Tian, Xingfa Gu, Yanting Zhang, Xuewen Li and Feng Zhang

Sensors 2025, 25(17), 5531; https://doi.org/10.3390/s25175531 - 5 Sep 2025

Cited by 2 | Viewed by 2344

Abstract

High-resolution remote sensing images often suffer from inadequate fusion between global and local features, leading to the loss of long-range dependencies and blurred spatial details, while also exhibiting limited adaptability to multi-scale object segmentation. To overcome these limitations, this study proposes RST-Net, a [...] Read more.

High-resolution remote sensing images often suffer from inadequate fusion between global and local features, leading to the loss of long-range dependencies and blurred spatial details, while also exhibiting limited adaptability to multi-scale object segmentation. To overcome these limitations, this study proposes RST-Net, a semantic segmentation network featuring a dual-branch encoder structure. The encoder integrates a ResNeXt-50-based CNN branch for extracting local spatial features and a Shunted Transformer (ST) branch for capturing global contextual information. To further enhance multi-scale representation, the multi-scale feature enhancement module (MSFEM) is embedded in the CNN branch, leveraging atrous and depthwise separable convolutions to dynamically aggregate features. Additionally, the residual dynamic feature fusion (RDFF) module is incorporated into skip connections to improve interactions between encoder and decoder features. Experiments on the Vaihingen and Potsdam datasets show that RST-Net achieves promising performance, with MIoU scores of 77.04% and 79.56%, respectively, validating its effectiveness in semantic segmentation tasks. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

16 pages, 9189 KB

Open AccessFeature PaperArticle

SEND: Semantic-Aware Deep Unfolded Network with Diffusion Prior for Multi-Modal Image Fusion and Object Detection

by Rong Zhang, Mao-Yi Xiong and Jun-Jie Huang

Mathematics 2025, 13(16), 2584; https://doi.org/10.3390/math13162584 - 12 Aug 2025

Cited by 3 | Viewed by 1610

Abstract

Multi-modality image fusion (MIF) aims to integrate complementary information from diverse imaging modalities into a single comprehensive representation and serves as an essential processing step for downstream high-level computer vision tasks. The existing deep unfolding-based processes demonstrate promising results; however, they often rely [...] Read more.

Multi-modality image fusion (MIF) aims to integrate complementary information from diverse imaging modalities into a single comprehensive representation and serves as an essential processing step for downstream high-level computer vision tasks. The existing deep unfolding-based processes demonstrate promising results; however, they often rely on deterministic priors with limited generalization ability and usually decouple from the training process of object detection. In this paper, we propose Semantic-Aware Deep Unfolded Network with Diffusion Prior (SEND), a novel framework designed for transparent and effective multi-modality fusion and object detection. SEND consists of a Denoising Prior Guided Fusion Module and a Fusion Object Detection Module. The Denoising Prior Guided Fusion Module does not utilize the traditional deterministic prior but combines the diffusion prior with deep unfolding, leading to improved multi-modal fusion performance and generalization ability. It is designed with a model-based optimization formulation for multi-modal image fusion, which is unfolded into two cascaded blocks: a Diffusion Denoising Fusion Block to generate informative diffusion priors and a Data Consistency Enhancement Block that explicitly aggregates complementary features from both the diffusion priors and input modalities. Additionally, SEND incorporates the Fusion Object Detection Module with the Denoising Prior Guided Fusion Module for object detection task optimization using a carefully designed two-stage training strategy. Experiments demonstrate that the proposed SEND method outperforms state-of-the-art methods, achieving superior fusion quality with improved efficiency and interpretability. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

25 pages, 3449 KB

Open AccessArticle

CSANet: Context–Spatial Awareness Network for RGB-T Urban Scene Understanding

by Ruixiang Li, Zhen Wang, Jianxin Guo and Chuanlei Zhang

J. Imaging 2025, 11(6), 188; https://doi.org/10.3390/jimaging11060188 - 9 Jun 2025

Cited by 1 | Viewed by 2090

Abstract

Semantic segmentation plays a critical role in understanding complex urban environments, particularly for autonomous driving applications. However, existing approaches face significant challenges under low-light and adverse weather conditions. To address these limitations, we propose CSANet (Context Spatial Awareness Network), a novel framework that [...] Read more.

Semantic segmentation plays a critical role in understanding complex urban environments, particularly for autonomous driving applications. However, existing approaches face significant challenges under low-light and adverse weather conditions. To address these limitations, we propose CSANet (Context Spatial Awareness Network), a novel framework that effectively integrates RGB and thermal infrared (TIR) modalities. CSANet employs an efficient encoder to extract complementary local and global features, while a hierarchical fusion strategy is adopted to selectively integrate visual and semantic information. Notably, the Channel–Spatial Cross-Fusion Module (CSCFM) enhances local details by fusing multi-modal features, and the Multi-Head Fusion Module (MHFM) captures global dependencies and calibrates multi-modal information. Furthermore, the Spatial Coordinate Attention Mechanism (SCAM) improves object localization accuracy in complex urban scenes. Evaluations on benchmark datasets (MFNet and PST900) demonstrate that CSANet achieves state-of-the-art performance, significantly advancing RGB-T semantic segmentation. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

25 pages, 12873 KB

Open AccessArticle

License Plate Detection Based on Improved YOLOv8n Network

by Ruizhe Zhu, Qiyang He, Hai Jin, Yonghua Han and Kejian Jiang

Electronics 2025, 14(10), 2065; https://doi.org/10.3390/electronics14102065 - 20 May 2025

Cited by 7 | Viewed by 5155

Abstract

To address the challenges of complex backgrounds, varying target scales, and dense targets in license plate detection within surveillance scenarios, we propose an enhanced method based on an improved YOLOv8n network. This approach involves redesigning key components of the YOLOv8n architecture, including the [...] Read more.

To address the challenges of complex backgrounds, varying target scales, and dense targets in license plate detection within surveillance scenarios, we propose an enhanced method based on an improved YOLOv8n network. This approach involves redesigning key components of the YOLOv8n architecture, including the C2f module, the SPPF module, and the detection head. Additionally, we optimize the WIoU loss function, replacing the original CIoU loss function, which leads to improved bounding box feature extraction and enhanced regression accuracy. To evaluate the model’s robustness in complex environments with varying lighting, backgrounds, angles, and vehicle types, we created a custom surveillance license plate dataset. Experimental results show that the improved model achieves a notable increase in detection accuracy, with mAP@0.5 rising from 90.9% in the baseline model to 94.4%, precision improving from 90.2% to 92.8%, and recall increasing from 82.9% to 87.9%. Additionally, the model’s parameters are reduced from 3.1 M to 2.1 M, significantly enhancing computational efficiency. Moreover, the model achieves an inference speed FPS of 86, maintaining high precision and meeting real-time detection requirements. This demonstrates that our method provides an efficient and reliable solution for license plate detection in surveillance scenarios. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

15 pages, 3571 KB

Open AccessArticle

Lightweight UAV Landing Model Based on Visual Positioning

by Ning Zhang, Junnan Tan, Kaichun Yan and Sang Feng

Sensors 2025, 25(3), 884; https://doi.org/10.3390/s25030884 - 31 Jan 2025

Cited by 7 | Viewed by 2307

Abstract

In order to enhance the precision of UAV (unmanned aerial vehicle) landings and realize the convenient and rapid deployment of the model to the mobile terminal, this study proposes a Land-YOLO lightweight UAV-guided landing algorithm based on the YOLOv8 n model. Firstly, GhostConv [...] Read more.

In order to enhance the precision of UAV (unmanned aerial vehicle) landings and realize the convenient and rapid deployment of the model to the mobile terminal, this study proposes a Land-YOLO lightweight UAV-guided landing algorithm based on the YOLOv8 n model. Firstly, GhostConv replaces standard convolutions in the backbone network, leveraging existing feature maps to create additional “ghost” feature maps via low-cost linear transformations, thereby lightening the network structure. Additionally, the CSP structure of the neck network is enhanced by incorporating the PartialConv structure. This integration allows for the transmission of certain channel characteristics through identity mapping, effectively reducing both the number of parameters and the computational load of the model. Finally, the bidirectional feature pyramid network (BiFPN) module is introduced, and the accuracy and average accuracy of the model recognition landing mark are improved through the bidirectional feature fusion and weighted fusion mechanism. The experimental results show that for the landing-sign data sets collected in real and virtual environments, the Land-YOLO algorithm in this paper is 1.4% higher in precision and 0.91% higher in mAP0.5 than the original YOLOv8n baseline, which can meet the detection requirements of landing signs. The model’s memory usage and floating-point operations per second (FLOPs) have been reduced by 42.8% and 32.4%, respectively. This makes it more suitable for deployment on the mobile terminal of a UAV. Full article

(This article belongs to the Topic Transformer and Deep Learning Applications in Image Processing)

► Show Figures

Figure 1

Topic Menu

Topic Editors

Transformer and Deep Learning Applications in Image Processing

Topic Information

Keywords

Participating Journals

Published Papers (14 papers)

Further Information

Guidelines

MDPI Initiatives

Follow MDPI

Topic Menu

Topic Editors

Transformer and Deep Learning Applications in Image Processing

Topic Information

Keywords

Participating Journals

Related Topic

Published Papers (14 papers)

Further Information

Guidelines

MDPI Initiatives

Follow MDPI