Macro placement is a crucial subproblem of chip design, focusing on determining the locations of numerous macros while minimizing multiple metrics. In recent years, reinforcement learning (RL) has gained traction as a favorable technique to improve placement performance. However, existing RL-based placers ignore the orientation of macros, resulting in the state space constrained to two-dimensional discrete coordinates and greatly restricting the exploration opportunities. To address this issue, we propose a novel macro placement method, RSPlace, which guides the bidirectional expansion of the global search tree to offer the RL agent more exploration opportunities, incorporating rotation into the RL-based macro placement solution for the first time. RSPlace intelligently determines the optimal rotation angle to maximize placement benefits by leveraging rotation sensing and placement perturbations. Extensive experiments demonstrate that taking the macro orientation into account substantially broadens the feasible locations and effectively reduces the half-perimeter wirelength (HPWL), thus ensuring that our approach significantly improves the optimization effect compared to the state-of-the-art method.

Camouflage Images Generation (CIG) is an emerging research area that focuses on synthesizing images in which objects are harmoniously blended and exhibit high visual consistency with their surroundings. Existing methods perform CIG by either fusing objects into specific backgrounds or outpainting the surroundings via foreground object-guided diffusion. However, they often fail to obtain natural results because they overlook the logical relationship between camouflaged objects and background environments. To address this issue, we propose CT-CIG, a Controllable Text-guided Camouflage Images Generation method that produces realistic and logically plausible camouflage images. Leveraging Large Visual Language Models (VLM), we design a Camouflage-Revealing Dialogue Mechanism (CRDM) to annotate existing camouflage datasets with high-quality text prompts. Subsequently, the constructed image-prompt pairs are utilized to finetune Stable Diffusion, incorporating a lightweight controller to guide the location and shape of camouflaged objects for enhanced camouflage scene fitness. Moreover, we design a Frequency Interaction Refinement Module (FIRM) to capture high-frequency texture features, facilitating the learning of complex camouflage patterns. Extensive experiments, including CLIPScore evaluation and camouflage effectiveness assessment, demonstrate the semantic alignment of our generated text prompts and CT-CIG's ability to produce photorealistic camouflage images.

Macro placement is a crucial sub-stage of chip design that focuses on determining the locations of numerous macros while optimizing multiple metrics. In recent years, evolutionary algorithms (EAs) have emerged as a promising technique for enhancing placement performance. However, the solution space of the previous methods is often limited by two-dimensional grids due to insufficient consideration of macro orientation during placement, leading to missed opportunities for further improvement. To address this issue, we propose EA-Rotation for macro placement, which incorporates rotation into the mutation operator to perturb the placement by adjusting the orientation of macros. Utilizing this efficient operator in conjunction with a greedy algorithm, EA-Rotation achieves superior placement solutions. Additionally, EA-Rotation serves as a fine-tuning technique for optimizing both the locations and orientation of macros, significantly improving the placement solutions of existing methods. Comprehensive experiments demonstrate that EA-Rotation outperforms state-of-the-art methods and is effectively integrated with other mutation operators.

Digital watermarking technology still faces significant challenges in real-world scenarios, primarily stemming from complex and unknown degradations that images may undergo during transmission and storage. Although current methods have made progress in specific degradation conditions by designing deep network architectures, their static models struggle to adapt to dynamically changing degradation environments. To address this, this paper proposes a dynamic degradation representation robust digital watermarking framework (DDRW). The core of this framework lies in establishing a bidirectional collaborative mechanism that integrates degradation perception and dynamic tuning. First, the DDRW designs a multi-mode dynamic composite attack noise layer. By leveraging hierarchical degradation space (partitioned into no/mild/moderate/severe attack subspaces) and a parameterized encoding mechanism, it randomly applies composite degradations with varying intensities and types during training, covering scenarios ranging from no attack to multistage severe degradations, thereby forcing the model to learn degradation-invariant features. Furthermore, a degradation representation network is introduced to quantify multi-dimensional degradation characteristics (e.g., blur and noise) in attacked images, generating degradation representation vectors. These vectors dynamically adjust parameter fusion weights in a multi-expert decoder, enabling adaptive collaborative optimization of watermark embedding strength and extraction strategies. Experimental results demonstrate that the DDRW achieves significantly higher watermark extraction accuracy under composite degradations compared to state-of-the-art deep learning-based watermarking baselines, thereby providing a novel technical solution for digital copyright protection in dynamic degradation environments.

The detection of traffic objects in aerial scenes holds significant application potential in both military and civilian sectors. However, current aerial traffic object detection techniques based on computer vision face challenges including limited awareness of object direction, a heavy computational burden on the feature extraction backbone network, and inadequate capacity to learn crucial semantic information. In this paper, our focus is on investigating the mechanisms for predicting the directional perception of traffic objects in aerial scenes, achieving backbone network lightness, and exploring methods for extracting key semantic information from objects. Firstly, to tackle the challenge of poor perception of traffic object direction and angle in aerial scenes, we utilize techniques like equivariant vector field convolution, multi-task anchor-free prediction, and adaptive loss to develop a precise mechanism for recognizing and predicting object directions. Secondly, given the presence of small-sized and numerous objects in aerial scenes, we propose the adoption of a lightweight backbone network employing channel stacking to decrease the model’s computational burden. Additionally, we establish a theoretical framework and methodology for optimizing and compressing this backbone network, aimed at enhancing feature extraction and propagation for aerial traffic objects. Furthermore, to address the issue of inadequate learning of key semantic information features, we incorporate saliency attention and multi-scale contextual information to capture the essential semantic characteristics of the objects. We also establish a method for extracting semantic features specifically for aerial traffic objects. The approach presented in this paper broadens the applicability of aerial object detection algorithms and offers novel methodologies and theoretical foundations for object detection in intricate scenarios.

Active learning (AL) and semi-supervised learning (SSL) both aim to reduce annotation costs: AL selectively annotates high-value samples from the unlabeled data, while SSL leverages abundant unlabeled data to improve model performance. Although these two appear intuitively compatible, directly combining them remains challenging due to fundamental differences in their frameworks. Current semi-supervised active learning (SSAL) methods often lack theoretical foundations and often design AL strategies tailored to a specific SSL algorithm rather than genuinely integrating the two fields. In this paper, we incorporate AL objectives into the overall risk formulation within the mainstream pseudo-label-based SSL framework, clarifying key differences between SSAL and traditional AL scenarios. To bridge these gaps, we propose a feature re-alignment module that aligns the features of unlabeled data under different augmentations by leveraging clustering and consistency constraints. Experimental results demonstrate that our module enables flexible combinations of SOTA methods from both AL and SSL, yielding more efficient algorithm performance.

Active learning (AL) is an effective method to balance annotation costs and model performance under resource-constrained circumstances. Most existing AL studies are typically designed for class-balanced datasets. However, the ubiquity of long-tailed distributions in real-world scenarios largely restricts the applicability of those AL methods. To tackle this problem, we propose a new active learning framework, namely long-tailed active learning (LTAL). The LTAL framework divides the long-tailed dataset into constantly evolving in-distribution (ID) and out-of-distribution (OOD) samples, and views the tail samples as OOD samples distinct from the head ones, thus intuitively converting the LTA problem into an iterative OOD detection task. We leverage an energy-based OOD detection approach with a well-designed class-imbalanced energy regularization loss to further extend the energy gap between head and tail classes, encouraging the model to select more unlabeled tail samples with higher free energy values. Experimental results show that despite its conceptual simplicity, the proposed method significantly outperforms competitive baselines.

Detecting camouflaged objects in underwater environments is crucial for marine ecological research and resource exploration. However, existing methods face two key challenges: underwater image degradation, including low contrast and color distortion, and the natural camouflage of marine organisms. Traditional image enhancement techniques struggle to restore critical features in degraded images, while camouflaged object detection (COD) methods developed for terrestrial scenes often fail to adapt to underwater environments due to the lack of consideration for underwater optical characteristics. To address these issues, we propose APGNet, an Adaptive Prior-Guided Network, which integrates a Siamese architecture with a novel prior-guided mechanism to enhance robustness and detection accuracy. First, we employ the Multi-Scale Retinex with Color Restoration (MSRCR) algorithm for data augmentation, generating illumination-invariant images to mitigate degradation effects. Second, we design an Extended Receptive Field (ERF) module combined with a Multi-Scale Progressive Decoder (MPD) to capture multi-scale contextual information and refine feature representations. Furthermore, we propose an adaptive prior-guided mechanism that hierarchically fuses position and boundary priors by embedding spatial attention in high-level features for coarse localization and using deformable convolution to refine contours in low-level features. Extensive experimental results on two public MAS datasets demonstrate that our proposed method APGNet outperforms 15 state-of-art methods under widely used evaluation metrics.

With the increasing size and complexity of software code, hidden defects can pose serious problems to systems, making zero‐defect software an urgent need for current industrial software applications. Software defect prediction (SDP) serves to identify defective modules or classes, with prediction models trained using historical defect data from various projects. This enables defect prediction in test projects, aiding in the rational allocation of test resources and the enhancement of software quality. The efficacy of SDP closely hinges on the quality of the defect dataset, the selected metric index, the trained model, and the algorithm design. This article reviews recent literature on SDP, summarizing existing research from three key perspectives: the dataset and metric elements employed in SDP, dataset optimization processing techniques, and defect prediction model techniques. It primarily focuses on introducing commonly used datasets and two types of defect metrics for SDP. Regarding dataset optimization processing technology, it discusses methods for handling abnormal data, high‐dimensional data, class imbalance data, and data disparity issues. Furthermore, it analyzes the construction of prediction models across four dimensions: supervised learning, semi‐supervised learning, unsupervised learning, and deep learning (DL). Key observations include: (i) Researchers utilize datasets of varying quality, performance evaluation metrics, and SDP models. The efficacy of software product metrics and development process metrics varies across different application scenarios, necessitating flexible metric selection based on actual requirements. (ii) Commonly used datasets like Promise and NASA exhibit varying data quality. Appropriate data preprocessing methods and dataset creation are crucial before training SDP models. (iii) In scenarios with limited labeled data, cross‐project transfer learning, semi‐supervised, or unsupervised learning methods tend to better utilize a broader range of training data. Given that each step in the SDP process corresponds to different unresolved issues, each requiring varying levels of response measures, we suggest that researchers comprehensively consider research objectives such as dataset quality, SDP model, performance evaluation indicators, and the need for model interpretability when conducting SDP‐related research. It’s important to note that no universal dataset or model can perform optimally across different application scenarios.

Using deep learning to detect road damage can significantly improve the effectiveness of road maintenance. To promote the development of road damage detection, we construct a high-resolution road damage data set named Asphalt Road Surface Disease Dataset(ARSDD), comprising 2297 images used in a real-world project. The annotation process is under the guidance of the road maintenance department, and it has more accurate labels and appropriate damage types. Most current road damage detection models are anchor-based, where one anchor corresponds to one sample. Therefore, these models are primarily limited by the setting of pre-defined anchors. Road damages have more extreme aspect ratios and scales than natural objects, and the general settings of anchors are inappropriate for road damage. In this paper, we propose a road damage detection model based on an improved adaptive training sample selection strategy, which can reduce manual anchor settings and is suitable for road damage detection. Moreover, as slight road damages tend to lose information during down-sampling, a cross-layer attention feature pyramid network is designed to compensate for this degradation in the spatial dimensions. While testing the ARSDD dataset, we find that the evaluation criterion for general object detection is unsuitable for road damage detection and propose a new post-processing method and diagonal-based evaluation criterion according to the characteristics of road damage. We validate our model using the 2018 Road Damage Dataset and our proposed dataset, and the results demonstrate the superiority of our model in road damage detection.Note to Practitioners—This paper was motivated by the problem of road damage detection, which is the key of intelligent road maintenance system. We comprehensively analyze the shortcomings of current publicly available content about road damage detection, including detection datasets, algorithms and evaluation metrics. Correspondingly, we first construct a road damage dataset from the real-world project, which contains six common categories of disease and is annotated under the guidance of the road maintenance department. Then we propose a detection model based on computer vision technology to improve the accuracy of damage detection. Finally, we propose a new post-processing method and a diagonal-based evaluation criterion based on the detection boxes. Any practitioner working on pavement inspection systems can use our released dataset and method to build a better-performing automated inspection system. In future research, we will try more refined inspection schemes and evaluation indicators, such as meshing the road surface images and evaluating the quality of the road based on the inspection results of each mesh.

Abstract Industrial image anomaly detection achieves automated detection and localization of defects or abnormal regions in images through image processing and deep learning techniques. Currently, utilizing the approach of reverse knowledge distillation has yielded favourable outcomes. However, it is still a challenge in terms of the feature extraction capability of the image and the robustness of the decoding of the student network. This study first addresses the issue that the teacher network has not been able to extract global information more effectively. To acquire more global information, a vision transformer network is introduced to enhance the model's global information extraction capability, obtaining better features to further assist the student network in decoding. Second, for anomalous samples, to address the low similarity between features extracted by the teacher network and features restored by the student network, Gaussian noise is introduced. This further increases the probability that the features decoded by the student model match normal sample features, enhancing the robustness of the student model. Extensive experiments were conducted on industrial image datasets AeBAD, MvtecAD, and BTAD. In the AeBAD dataset, under the PRO performance metric, the result is 89.83%, achieving state‐of‐the‐art performance. Under the AUROC performance metric, it reaches 83.35%. Similarly, good results were achieved on the MvtecAD and BTAD datasets. The proposed method's effectiveness and performance advantages were validated across multiple industrial datasets, providing a valuable reference for the application of industrial image anomaly detection methods.

Compared to CNN-based methods, Transformer-based methods achieve impressive image restoration outcomes due to their ability to model remote dependencies. However, how to apply Transformer-based methods to the field of blind super-resolution (SR) and further make an SR network adaptive to degradation information is still an open problem. In this paper, we propose a new degradation-aware self-attention-based Transformer model, where we incorporate contrastive learning into the Transformer network for learning the degradation representations of input images with unknown noise. In particular, we integrate both CNN and Transformer components into the SR network, where we first use the CNN modulated by the degradation information to extract local features, and then employ the degradation-aware Transformer to extract global semantic features. We apply our proposed model to several popular large-scale benchmark datasets for testing, and achieve the state-of-the-art performance compared to existing methods. In particular, our method yields a PSNR of 32.43 dB on the Urban100 dataset at ×2 scale, 0.94 dB higher than DASR, and 26.62 dB on the Urban100 dataset at ×4 scale, 0.26 dB improvement over KDSR, setting a new benchmark in this area. The source code is available at:https://github.com/I2-Multimedia-Lab/DSAT/tree/main.

Active learning (AL) aims to find a better trade-off between labeling costs and model performance by consciously selecting more informative samples to label. Recently, adversarial approaches have emerged as effective solutions. Most of them leverage generative adversarial networks to align feature distributions of labeled and unlabeled data, upon which discriminators are trained to better distinguish between them. However, these methods fail to consider the relationship between unlabeled samples and decision boundaries, and their training processes are often complex and unstable. To this end, this paper proposes a novel adversarial AL method, namely multi-classifier adversarial optimization for active learning (MAOAL). MAOAL employs task-specific decision boundaries for data alignment while selecting the most informative samples to label. To fulfill this, we introduce a novel classifier class confusion (C3) metric, which represents the classifier discrepancy as the inter-class correlation of classifier outputs. Without any additional hyper-parameters, the C3 metric further reduces the negative impacts of ambiguous samples in the process of distribution alignment and sample selection. More concretely, the network is trained adversarially by adding two auxiliary classifiers, reducing the distribution bias of labeled and unlabeled samples by minimizing the C3 loss between classifiers, while learning tighter decision boundaries and highlighting hard samples by maximizing the C3 loss. Finally, the unlabeled samples with the highest C3 loss are selected to label. Extensive experiments demonstrate the superiority of our approach over state-of-the-art AL methods in terms of image classification and object detection.

Scene text recognition (STR) has been actively studied in recent years, with a wide range of applications in autonomous driving, image retrieval and much more. However, when a pre-trained deep STR model learns a new task, its performance on previous tasks may drop dramatically, due to catastrophic forgetting in deep neural networks. A potential solution to combat the forgetting of prior knowledge is incremental learning (IL), which has shown its effectiveness and significant progress in image classification. Yet, exploiting IL in the context of STR has been barely visited, probably because the forgetting problem is even worse in STR. To address this issue, we propose the lifelong scene text recognizer (LSTR) that learns STR tasks incrementally while alleviating forgetting. Specifically, LSTR assigns each task a set of task-specific expert modules at different stages of an STR model, while other parameters are shared among tasks. These shared parameters are only learned in the first task and remain unchanged during subsequent learning to ensure that no learned knowledge is overlooked. Moreover, in real applications, there is no prior knowledge about which task an input image belongs to, making it impossible to precisely select the corresponding expert modules. To this end, we propose the incremental task prediction network (ITPN) to identify the most related task category by pulling the features of the same task closer and pushing those of different tasks farther apart. To validate the proposed method in our newly-introduced IL setting, we collected a large-scale dataset consisting of both real and synthetic multilingual STR data. Extensive experiments on this dataset clearly show the superiority of our LSTR over state-of-the-art IL methods.

Due to the more dramatic multi-scale variations and more complicated foregrounds and backgrounds in optical remote sensing images (RSIs), the salient object detection (SOD) for optical RSIs becomes a huge challenge. However, different from natural scene images (NSIs), the discussion on the optical RSI SOD task still remains scarce. In this paper, we propose a multi-scale context network, namely MSCNet, for SOD in optical RSIs. Specifically, a multi-scale context extraction module is adopted to address the scale variation of salient objects by effectively learning multi-scale contextual information. Meanwhile, in order to accurately detect complete salient objects in complex backgrounds, we design an attention-based pyramid feature aggregation mechanism for gradually aggregating and refining the salient regions from the multi-scale context extraction module. Extensive experiments on two benchmarks demonstrate that MSCNet achieves competitive performance with only 3.26M parameters. The code will be available at https://github.com/NuaaYH/MSCNet.

Vehicle detection in aerial photography scenarios has a wide range of promising applications in the military and civilian fields. Recently, object detection algorithms based on depth models have shown superior performance in aerial vehicle detection tasks. However, these detection algorithms are often accompanied by a large amount of computation and resource consumption, which leads to the inability to perform real-time detection. In addition, the insufficient feature extraction capability of the vehicle and the complex background information also lead to low detection accuracy. In this letter, we propose a lightweight backbone network with a context information module and an attention mechanism module for vehicle detection in the aerial image, which enables the feature extraction network to increase the utilization of contextual information and salient regions. In addition, we use adaptive anchor-free in the detection model to predict the bounding box. The proposed detection algorithm achieves 89.7% and 94.1% mean average precision (mAP) on the German Aerospace Center (DLR)-3K dataset and the created dataset, and the detection time for each image is 1.66 and 0.049 s, respectively.

The widespreadPomacea canaliculataandPomacea maculatain North America and Asia have caused significant adverse effects on the local ecological environment and residents’ health. Timely knowledge of the distribution of eggs from the twoPomaceaspp. in a certain region can effectively reduce the cost of treatment and improve prevention effectiveness. Most of the existing methods are only able to identify eggs from the twoPomaceaspp. or detect them in specific but not natural environments while they cannot achieve good results in the face of a complex real-world scene. This letter proposes a model for detecting eggs from the twoPomaceaspp. based on dynamic convolution and multiscale feature fusion. The model can identify and locate the eggs of the twoPomaceaspp. effectively. At the same time, we combined the proposed model with scale invariant feature transform (SIFT) algorithm to design a system for counting eggs of the twoPomaceaspp., which can automatically identify the eggs in the actual natural environment and alleviate duplicate counting caused by image acquisition. Besides, we also built a dataset of 20 000 images ofPomacea canaliculataeggs andPomacea maculataeggs from unmanned aerial vehicle (UAV) aerial photography. Experimental results showed that the proposed deep learning model has a better performance than others, and the proposed computer vision system can be successfully applied to supportPomaceaspp. disease management.

Vehicle detection in the aerial image is an essential and challenging task widely used in industry and agriculture. Deep learning technology has recently achieved rapid development and good object detection results. However, the background of aerial images is complex, targets are densely distributed, and some of them are occluded. For densely distributed targets, we need to predict at each feature point. In the case of complex background and target occlusion, it is often difficult to determine whether a location contains a target if the model only focuses on the local information. Therefore, we need a global perspective and contextual information to help train the model. This paper proposes a new anchor-free small object detection algorithm, which improves feature extraction by fusing contextual semantic information. In addition, a dynamic activation function is also used in our network, which helps us calculate the activation function value for each point from a global perspective. Moreover, we also use the channel attention module and the transformer as the spatial attention module to help the network efficiently obtain global information. We evaluate the effectiveness of our method on the public dataset DLR-3K and VEDAI, and the AP achieves 0.896 and 0.875.

Vehicle object detection in aerial scenes has important applications in both military and civilian fields. Recently, deep learning has shown clear advantages in object detection, and the detection performance has been continuously improved. However, these deep object detection algorithms rely on anchor-based approaches accompanied by complex convolutional operations. In this paper, we establish a lightweight aerial vehicle object detection algorithm based on the method of anchor-free. The anchor-free based object detection method effectively gets rid of the limitation of detection model capability by the size of fixed anchor box, which reduces the set of parameters and provides a more flexible solution space. In addition, the proposed lightweight object feature extraction network effectively reduces the computational cost of the model, while improving the feature extraction capability of small objects. Besides, we use channel stacking to improve the object feature extraction capability of the lightweight network, and introduce the attention mechanism in the detection model to improve the efficiency of resource utilization. We evaluate the proposed detection algorithm on both the public aerial dataset and our collected aerial dataset, and the results show that our algorithm has significant advantages over other detection algorithms in detection accuracy and efficiency. The proposed detection algorithm achieves 89.1% and 92.6% mAP on the Munich dataset and the created dataset, and the detection time for each image is 1.21s and 0.036s, respectively.

The appearance degradation caused by low resolution is the core problem of small face detection. Therefore, a natural approach is to assemble information from the context. This paper focuses on how to use high-level contextual information to improve the abilities of anchor-based detectors to detect dense and degenerate faces. We tap the spatial contextual information on the overall view based on the density map, and propose the prior of face co-occurrence for inferred bounding-boxes coordination. We also propose score-size-specific non-maximum suppression to replace the traditional non-maximum suppression at the end of anchor-based detectors. According to the inferred face boxes' quantity, score and size, the proposed synthetical solution reduces false positives and increases true positives. Our method does not require additional training, which is model-independent and can be embedded into existing face detectors. We also propose a dataset - Crowd Face for face detection, which is full of challenges. We expect to supply enough samples to highlight the difficulties of detecting dense and degenerate faces. We embed our proposed methods into state-of-the-art face detectors on massively benchmarked face datasets. Compared with the prior art on the WIDER FACE hard set, our method increase an Average Precision of 0.1 %-1.3%. On Crowd Face, it increases an Average Precision of 1 % – 6%. Dataset is available on: https://github.com/QxGeng/Crowd-Face.

Modern supervised learning relies on a large amount of training data, yet there are many noisy annotations in real datasets. For semantic segmentation tasks, pixel-level annotation noise is typically located at the edge of an object, while pixels within objects are fine-annotated. We argue the coarse annotations can provide instructive supervised information to guide model training rather than be discarded. This paper proposes a noise learning framework based on knowledge distillation NLKD, to improve segmentation performance on unclean data. It utilizes a teacher network to guide the student network that constitutes the knowledge distillation process. The teacher and student generate the pseudo-labels and jointly evaluate the quality of annotations to generate weights for each sample. Experiments demonstrate the effectiveness of NLKD, and we observe better performance with boundary-aware teacher networks and evaluation metrics. Furthermore, the proposed approach is model-independent and easy to implement, appropriate for integration with other tasks and models.

Using only one deep model for cross scene video foreground segmentation is still very challenging because existing methods are scene-dependent, which restricts the consistent segmentation. In this paper, we propose a cross scene video foreground segmentation framework to extend the generalization capability of those supervised model depending on scene-specific training. The proposed framework flexibly utilizes three well-trained supervised models as guidance to yield a coarse segmentation mask. The co-occurrence probability-based unsupervised background subtraction model is introduced to achieve scene adaptation in the plug and play style without any fine-tuning and labels. Experimental results on LIMU and CDNet2014 datasets validate our framework outperforms the state-of-the-art supervised/unsupervised approaches that participate in the comparison. Experiments also show the training efficiency-related improvements – when introducing the guidance models, the demand for quantity and quality of training samples to train the unsupervised model is reduced. Codes https://github.com/MeteoorLiu/Venus/tree/MeteoorLiu-SUMC

Abstract Vehicle detection in aerial images is an interesting and challenging task. Traditional methods are based on sliding‐window search and handcrafted features, which limits the representation power and has heavy computational costs. Recent research have shown that deep‐learning algorithms are widely used in the field of object detection. However, the deep‐learning algorithms still face many difficulties and challenges in the object detection under the aerial scene. Meanwhile, the high computational cost of detection algorithms lead to low‐detection efficiency. In this study, we build a fast and accurate lightweight detection framework for vehicle detection in aerial scenes. The proposed detection method improves the expressive ability of detection network and significantly reduces the amount of calculations in the model. Meanwhile, setting suitable anchor boxes according to the size of the object vehicles have been introduced in our model, which also effectively improves the performance of the detection. In addition, we have published a new aerial vehicle image dataset and verified the effectiveness of our method. In the Munich dataset and our dataset, our method achieves 85.8% and 91.2% of the mean average precision (mAP), and its detection time is 1.78 and 0.048 s on Nvidia Titan XP. Our results show that the proposed framework achieves significant improvement over several alternatives and state‐of‐the‐art schemes with higher accuracy and less detection time.

Abstract The semantic representation of deep features is essential for image context understanding, and effective fusion of features with different semantic representations can significantly improve the model's performance on salient object detection. This paper proposes a novel method called multi‐receptive and parallel integration, for salient object detection. Firstly, a multi‐receptive enhancement module is designed to effectively expand the receptive fields of features from different layers and generate features with different receptive fields. Multi‐receptive enhancement module can enhance the semantic representation and improve the model's perception of the image context, which enables the model to locate the salient object accurately. Secondly, in order to reduce the reuse of redundant information in the complex top‐down fusion method and weaken the differences between semantic features, a relatively simple but effective parallel fusion strategy is proposed. It allows multi‐scale features to better interact with each other, thus improving the overall performance of the model. Experimental results on multiple datasets demonstrate that the proposed method outperforms state‐of‐the‐art methods under different evaluation metrics.

With the rapid development of the electronic industry, the defect detection of printed circuit board (PCB) components is becoming more and more important. The types of PCB components are diverse and accompanied by complex character information, which is difficult to identify. The traditional detection method is inefficient, and it is unable to effectively perform the diversified category detection of PCB components and character recognition in complex scenes. The deep convolutional neural network has obvious advantages in object detection and character recognition, which can be used to implement a PCB component defect detection system. In this study, the authors have established a lightweight PCB type detection model called LD‐PCB, which can perform real‐time detection while improving detection accuracy. In addition, in the character detection of PCB, they have established a fast and robust character recognition model, called CR‐PCB. This model can effectively improve the accuracy of irregular character recognition. Finally, they established and published a dataset of PCB components, and combined with LD‐PCB and CR‐PCB to realise the PCB defect detection system. This system can realise the functions of defect detection, wrong insertion, missing insertion, and character recognition in industrial PCB production. The results show that the method proposed in this study can effectively detect defects on PCB components.