Intra-class Variations Research Articles

EC-PFN: a multiscale woven fusion network for industrial product surface defect detection

In order to address challenges such as small target sizes, low contrast, significant intra-class variations, and indistinct inter-class differences in surface defect detection, this paper proposes the Enhanced Context-aware Parallel Fusion Network (EC-PFN). The network employs a Featur Weave Network architecture to enhance contextal awareess and parallel fusion capabilities. It utilizes a Feature Fusion Module (UniFusionLayer) for effective multiscale and multisemantic feature learning, offering new perspectives on feature fusion. Additionally, a Receptive Field Block (RFB) module is introduced to expand the receptive field, enhancing feature extraction in scenarios with low contrast and subtle defects. The Loss Ranking Module (LRM) is incorporated to optimize the target-oriented loss, improving performance by omitting low-confidence bounding boxes. Extensive experiments on a light guide plate defect dataset demonstrate that EC-PFN achieves a detection accuracy (mAP) of 98.9%, a detection speed of 92 FPS, and a computational cost of 14.5 GFLOPs, outperforming mainstream surface defect detection models.

OMAD-6: Advancing Offshore Mariculture Monitoring with a Comprehensive Six-Type Dataset and Performance Benchmark

Offshore mariculture is critical for global food security and economic development. Advances in deep learning and data-driven approaches, enable the rapid and effective monitoring of offshore mariculture distribution and changes. However, detector performance depends heavily on training data quality. The lack of standardized classifications and public datasets for offshore mariculture facilities currently hampers effective monitoring. Here, we propose to categorize offshore mariculture facilities into six types: TCC, DWCC, FRC, LC, RC, and BC. Based on these categories, we introduce a benchmark dataset called OMAD-6. This dataset includes over 130,000 instances and more than 16,000 high-resolution remote sensing images. The images with a spatial resolution of 0.6 m were sourced from key regions in China, Chile, Norway, and Egypt, from the Google Earth platform. All instances in OMAD-6 were meticulously annotated manually with horizontal bounding boxes and polygons. Compared to existing remote sensing datasets, OMAD-6 has three notable characteristics: (1) it is comparable to large, published datasets in instances per category, image quantity, and sample coverage; (2) it exhibits high inter-class similarity; (3) it shows significant intra-class diversity in facility sizes and arrangements. Based on the OMAD-6 dataset, we evaluated eight state-of-the-art methods to establish baselines for future research. The experimental results demonstrate that the OMAD-6 dataset effectively represents various real-world scenarios, which have posed considerable challenges for current instance segmentation algorithms. Our evaluation confirms that the OMAD-6 dataset has the potential to improve offshore mariculture identification. Notably, the QueryInst and PointRend algorithms have distinguished themselves as top performers on the OMAD-6 dataset, robustly identifying offshore mariculture facilities even with complex environmental backgrounds. Its ongoing development and application will play a pivotal role in future offshore mariculture identification and management.

Mine Your Own Anatomy: Revisiting Medical Image Segmentation With Extremely Limited Labels.

Recent studies on contrastive learning have achieved remarkable performance solely by leveraging few labels in the context of medical image segmentation. Existing methods mainly focus on instance discrimination and invariant mapping (i.e., pulling positive samples closer and negative samples apart in the feature space). However, they face three common pitfalls: (1) tailness: medical image data usually follows an implicit long-tail class distribution. Blindly leveraging all pixels in training hence can lead to the data imbalance issues, and cause deteriorated performance; (2) consistency: it remains unclear whether a segmentation model has learned meaningful and yet consistent anatomical features due to the intra-class variations between different anatomical features; and (3) diversity: the intra-slice correlations within the entire dataset have received significantly less attention. This motivates us to seek a principled approach for strategically making use of the dataset itself to discover similar yet distinct samples from different anatomical views. In this paper, we introduce a novel semi-supervised 2D medical image segmentation framework termed Mine yOur owNAnatomy (MONA), and make three contributions. First, prior work argues that every pixel equally matters to the model training; we observe empirically that this alone is unlikely to define meaningful anatomical features, mainly due to lacking the supervision signal. We show two simple solutions towards learning invariances-through the use of stronger data augmentations and nearest neighbors. Second, we construct a set of objectives that encourage the model to be capable of decomposing medical images into a collection of anatomical features in an unsupervised manner. Lastly, we both empirically and theoretically, demonstrate the efficacy of our MONA on three benchmark datasets with different labeled settings, achieving new state-of-the-art under different labeled semi-supervised settings. MONA makes minimal assumptions on domain expertise, and hence constitutes a practical and versatile solution in medical image analysis. We provide the PyTorch-like pseudo-code in supplementary.

Unsupervised prototype self-calibration based on hybrid attention contrastive learning for enhanced few-shot action recognition

The collection and annotation of large-scale video data pose significant challenges, prompting the exploration of few-shot models to recognize unseen actions with limited training samples. However, most existing models rely on complex temporal alignment strategies, neglecting the necessity of constructing accurate class prototypes in action recognition. To address this limitation, we propose an unsupervised prototype self-calibration based on hybrid attention contrastive learning (UPSHC), aiming to refine prototypes through various strategies. Our network comprises three innovative components: a hybrid attention network (HAN), a dual adaptive contrastive learning mechanism (DACL), and an unsupervised prototype self-calibration mechanism (UPSC). The HAN integrates self-attention and cross-attention mechanisms to highlight important spatiotemporal regions, reducing intra-class variations and enhancing inter-class variations. The DACL mechanism constructs prototype-centered and query-centered contrastive losses to deeply explore the similarities between prototypes and samples, improving the model’s sensitivity to subtle inter-class differences. Additionally, DACL employs adaptive margins to dynamically adjust the distances between positive and negative samples, mitigating the unreliability of fixed margins. The UPSC module utilizes unlabeled query samples in an unsupervised manner to optimize target prototypes without requiring additional training. Experimental results on three benchmark datasets demonstrate that UPSHC achieves state-of-the-art performance, highlighting the superiority of our prototype enhancement approach in few-shot action recognition.

Domain-Invariant Few-Shot Contrastive Learning for Hyperspectral Image Classification

In Hyperspectral Image (HSI) classification, acquiring large quantities of high-quality labeled samples is typically costly and impractical. Traditional deep learning methods are limited in such scenarios due to their dependence on sample quantities. To address this challenge, researchers have turned to Few-Shot Learning (FSL). Although existing FSL methods improve classification performance by enhancing domain invariance through domain adaptation, they often overlook the critical issue of high inter-class similarity and large intra-class variability. Moreover, during domain alignment, features of different categories within the same domain may become confused. To address these issues, this paper proposes a novel Domain-Invariant Few-Shot Contrastive Learning (DIFSCL) method, which combines domain adaptation and contrastive learning strategies to not only learn domain-invariant features but also significantly enhance inter-class discriminability. Based on this, we further design a multi-scale adaptive attention mechanism for a hyperspectral feature extraction network to more effectively extract and optimize generalized features within the DIFSCL framework, significantly improving intra-class consistency and inter-class discriminative capability of features. Experimental results on three widely used HSI datasets demonstrate that our method significantly outperforms existing techniques in few-shot classification tasks.

ReMamba: a hybrid CNN-Mamba aggregation network for visible-infrared person re-identification.

Visible-Infrared Person Re-identification (VI-ReID) has been consistently challenged by the significant intra-class variations and cross-modality differences between different cameras. Therefore, the key lies in how to extract discriminative modality-shared features. Existing VI-ReID methods based on Convolutional Neural Networks (CNN) and Vision Transformers (ViT) have shortcomings in capturing global features and controlling computational complexity, respectively. To tackle these challenges, we propose a hybrid network framework called ReMamba. Specifically, we first use a CNN as the backbone network to extract multi-level features. Then, we introduce the Visual State Space (VSS) model, which is responsible for integrating the local features output by the CNN from lower to higher levels. These local features serve as a complement to global information and thereby enhancing the local details clarity of the global features. Considering the potential redundancy and semantic differences between local and global features, we design an adaptive feature aggregation module that automatically filters and effectively aggregates both types of features, incorporating an auxiliary aggregation loss to optimize the aggregation process. Furthermore, to better constrain cross-modality features and intra-modal features, we design a modal consistency identity constraint loss to alleviate cross-modality differences and extract modality-shared information.Extensive experiments conducted on the SYSU-MM01, RegDB, and LLCM datasets demonstrate that our proposed ReMamba outperforms state-of-the-art VI-ReID methods.

Edge-guided and hierarchical aggregation network for robust medical image segmentation

Medical image segmentation with the convolutional neural networks (CNNs), significantly enhances clinical analysis and disease diagnosis. However, medical images inherently exhibit large intra-class variability, minimal inter-class differences, and substantial noise levels. Extracting robust contextual information and aggregating discriminative features for fine-grained segmentation remains a formidable task. Additionally, existing methods often struggle with producing accurate mask edges, leading to blurred boundaries and reduced segmentation precision. This paper introduces a novel Edge-guided and Hierarchical Aggregation Network (EHANet) which excels at capturing rich contextual information and preserving fine spatial details, addressing the critical issues of inaccurate mask edges and detail loss prevalent in current segmentation models. The Inter-layer Edge-aware Module (IEM) enhances edge prediction accuracy by fusing early encoder layers, ensuring precise edge delineation. The Efficient Fusion Attention Module (EFA) adaptively emphasizes critical spatial and channel features while filtering out redundancies, enhancing the model’s perception and representation capabilities. The Adaptive Hierarchical Feature Aggregation Module (AHFA) module optimizes feature fusion within the decoder, maintaining essential information and improving reconstruction fidelity through hierarchical processing. Quantitative and qualitative experiments on four public datasets demonstrate the effectiveness of EHANet in achieving superior mIoU, mDice, and edge accuracy against eight other state-of-the-art segmentation methods, highlighting its robustness and precision in diverse clinical scenarios.

Multimodal Information Fusion Dynamic Target Recognition for Autonomous Driving

With the continuous development of autonomous driving technology, realizing high-precision road obstacle detection is crucial to ensure traffic safety and driving experience. However, traditional obstacle detection methods often perform poorly in complex driving scenarios, such as obstacle movement and occlusion. To cope with this problem, this study proposes a road obstacle detection method based on a two-stream convolutional neural network model, aiming to overcome the limitations of traditional methods in capturing spatiotemporal features and handling complex situations. Our research approach is based on the following innovations: First, we introduce a dual-stream convolutional neural network structure, where one stream is used to extract spatial features from the contour information of the obstacle frames, and the other stream extracts temporal features from the temporal stream information. This dual-stream structure can fully capture the appearance and dynamic information of the obstacles, thus improving detection accuracy. Second, we design a feature fusion module to fuse the two features to obtain richer obstacle features. In addition, we propose a new loss function, i.e. clustering loss, for better optimizing the model training process, reducing intra-class variation, and increasing inter-class differences, thus improving the generalization performance of the model. In the experimental section, we conducted extensive experimental analysis using Citycapes and BDD100K datasets. The experimental results show that our model achieves significant performance improvement compared to both the traditional convolutional neural network method and the YOLOv5 method in a variety of scenarios such as obstacle stationary, obstacle moving, and obstacle with occlusion. Specifically, our method improves the recognition rate up to 4.8% to 14.5% on the Citycapes dataset and 6.5% to 12.8% on the BDD100K dataset, respectively, under different scenarios. In addition, our model also exhibits more advantages on small datasets, showing higher generalization ability and robustness.

MP-GIEN: Vehicle Re-Identification Method Based on Multi-View Progressive Graph Interactive Embedding Network

Vehicle re-identification primarily faces two challenges: significant intra-class variations caused by different camera views and subtle inter-class differences between vehicles of the same model. In this paper, we propose a Multi-View Progressive Graph Interaction Embedding Network (MP-GIEN) for vehicle re-identification. First, we design a global semantic extractor. By employing Graph Interaction Units (GI Units) and semantic context, we combine semantic information to facilitate contextual reasoning of image regions, thereby extracting global semantic features of the target. The global loss is then calculated using ID loss and triplet loss. Second, we devise a local view-aware extractor. The U-Net network is utilized to parse the image into four views (front, rear, top, and side). Subsequently, features are aligned through mask average pooling, and a progressive training strategy is adopted to address the learning of fine-grained information. Lastly, we design a feature fusion enhancer. We revise the typical triplet loss to avoid mismatches in local features. By optimizing the local triplet loss and global loss, we learn the embedding of visual features, which not only reduces the intra-instance distance but also enlarges the inter-instance differences. MP-GIEN facilitates the capture of stable discriminative information on vehicles under different views. Experiments conducted on the VeRi-Wild dataset demonstrate that our model significantly outperforms state-of-the-art methods.

Category-Level Pose Estimation and Iterative Refinement for Monocular RGB-D Image

Category-level pose estimation is proposed to predict the 6D pose of objects under a specific category and has wide applications in fields such as robotics, virtual reality, and autonomous driving. With the development of VR/AR technology, pose estimation has gradually become a research hotspot in 3D scene understanding. However, most methods fail to fully utilize geometric and color information to solve intra-class shape variations, which leads to inaccurate prediction results. To solve the above problems, we propose a novel pose estimation and iterative refinement network, use an attention mechanism to fuse multi-modal information to obtain color features after a coordinate transformation, and design iterative modules to ensure the accuracy of object geometric features. Specifically, we use an encoder-decoder architecture to implicitly generate a coarse-grained initial pose and refine it through an iterative refinement module. In addition, due to the differences between rotation and position estimation, we design a multi-head pose decoder that utilizes the local geometry and global features. Finally, we design a transformer-based coordinate transformation attention module to extract pose-sensitive features from RGB images and supervise color information by correlating point cloud features in different coordinate systems. We train and test our network on the synthetic dataset CAMERA25 and the real dataset REAL275. Experimental results show that our method achieves state-of-the-art performance on multiple evaluation metrics.

Entropy-Based Ensemble of Convolutional Neural Networks for Clothes Texture Pattern Recognition

Automatic clothes pattern recognition is important to assist visually impaired people and for real-world applications such as e-commerce or personal fashion recommendation systems, and it has attracted increased interest from researchers. It is a challenging texture classification problem in that even images of the same texture class expose a high degree of intraclass variations. Moreover, images of clothes patterns may be taken in an unconstrained illumination environment. Machine learning methods proposed for this problem mostly rely on handcrafted features and traditional classification methods. The research works that utilize the deep learning approach result in poor recognition performance. We propose a deep learning method based on an ensemble of convolutional neural networks where feature engineering is not required while extracting robust local and global features of clothes patterns. The ensemble classifier employs a pre-trained ResNet50 with a non-local (NL) block, a squeeze-and-excitation (SE) block, and a coordinate attention (CA) block as base learners. To fuse the individual decisions of the base learners, we introduce a simple and effective fusing technique based on entropy voting, which incorporates the uncertainties in the decisions of base learners. We validate the proposed method on benchmark datasets for clothes patterns that have six categories: solid, striped, checkered, dotted, zigzag, and floral. The proposed method achieves promising results for limited computational and data resources. In terms of accuracy, it achieves 98.18% for the GoogleClothingDataset and 96.03% for the CCYN dataset.

Knowledge Guided Transformer Network for Compositional Zero-Shot Learning

Compositional Zero-shot Learning (CZSL) attempts to recognise images of new compositions of states and objects when images of only a subset of state-object compositions are available as training data. An example of CZSL is to recognise images of peeled apple by a model when it is trained using images of peeled orange , ripe apple and ripe orange . There are two major challenges in solving CZSL. First, the visual features of a state vary depending on the context of a state-object composition. For example state like ripe produces distinct visual properties in the compositions ripe orange and ripe banana . Hence, understanding the context dependency of state features is a necessary requirement to solve CZSL. Second, the extent of association between the features of a state and an object varies significantly in different images of same composition. For example, in different images of peeled oranges , the oranges may be peeled to different extents. As a consequence, the visual features of images of the class peeled orange may vary. Hence, there exists a significant amount of intra-class variability among the visual features of different images of a composition. Existing approaches merely look for the existence or absence of features of particular state or object in a composition. Our approach not only looks for the existence of a particular state features or object features but also the extent of association of state features and object features to better tackle the intra-class variability in visual features of compositional images. The proposed architecture is constructed using a novel Knowledge Guided Transformer . The transformer-based framework is utilised for processing larger context dependency between the state and object. Extensive experiments on C-GQA, MIT-States and UT-Zappos50k datasets demonstrate the superiority of the proposed approach in comparison with the state-of-the-art in both open-world and closed-world CZSL settings.

Attention-disentangled re-ID network for unsupervised domain adaptive person re-identification

Unsupervised domain adaptation (UDA) for person re-identification (re-ID) aims to bridge the domain gap by transferring knowledge from the labeled source domain to the unlabeled target domain. Recently, pseudo-label-based approaches have become the dominant solution for addressing this issue. However,most pseudo-label-based methods neglect class boundary samples to reduce the influence of false pseudo labels, sacrificing intra-class semantic diversity. Although hard sample memory bank-based approaches can discover relationships between samples and describe intra-class diversity, they are prone to increasing the risks of generating incorrect pseudo labels. Balancing intra-class variety and pseudo-label accuracy in cross-domain person re-ID poses a challenge. In this study, we propose an attention-disentangled re-ID network (ADDNet) to enhance the discriminative ability of re-ID related feature representations, addressing the contradiction between intra-class diversity and pseudo-label accuracy. Unlike most disentanglement learning-based re-ID approaches that focus on separating explanatory factors, we design a spatial attention-disentangled mechanism to separate re-ID related and unrelated weights, enhancing the discriminative ability of cross-domain feature representation. Additionally, a multiple hard sample memory learning strategy is designed to express intra-class diversity of target samples using re-ID related features extracted from ADDNet. Extensive experiments show that ADDNet achieves 46.7% mAP on the Market-to-MSMT cross-domain re-ID task, outperforming state-of-the-art methods by 6.5 points.

Prototype Correlation Matching and Class-Relation Reasoning for Few-Shot Medical Image Segmentation.

Few-shot medical image segmentation has achieved great progress in improving accuracy and efficiency of medical analysis in the biomedical imaging field. However, most existing methods cannot explore inter-class relations among base and novel medical classes to reason unseen novel classes. Moreover, the same kind of medical class has large intra-class variations brought by diverse appearances, shapes and scales, thus causing ambiguous visual characterization to degrade generalization performance of these existing methods on unseen novel classes. To address the above challenges, in this paper, we propose a Prototype correlation Matching and Class-relation Reasoning (i.e., PMCR) model. The proposed model can effectively mitigate false pixel correlation matches caused by large intra-class variations while reasoning inter-class relations among different medical classes. Specifically, in order to address false pixel correlation match brought by large intra-class variations, we propose a prototype correlation matching module to mine representative prototypes that can characterize diverse visual information of different appearances well. We aim to explore prototypelevel rather than pixel-level correlation matching between support and query features via optimal transport algorithm to tackle false matches caused by intra-class variations. Meanwhile, in order to explore inter-class relations, we design a class-relation reasoning module to segment unseen novel medical objects via reasoning inter-class relations between base and novel classes. Such inter-class relations can be well propagated to semantic encoding of local query features to improve few-shot segmentation performance. Quantitative comparisons illustrates the large performance improvement of our model over other baseline methods.

An Adaptive Deep Metric Learning Loss Function for Class-Imbalance Learning via Intraclass Diversity and Interclass Distillation.

Deep metric learning (DML) has been widely applied in various tasks (e.g., medical diagnosis and face recognition) due to the effective extraction of discriminant features via reducing data overlapping. However, in practice, these tasks also easily suffer from two class-imbalance learning (CIL) problems: data scarcity and data density, causing misclassification. Existing DML losses rarely consider these two issues, while CIL losses cannot reduce data overlapping and data density. In fact, it is a great challenge for a loss function to mitigate the impact of these three issues simultaneously, which is the objective of our proposed intraclass diversity and interclass distillation (IDID) loss with adaptive weight in this article. IDID-loss generates diverse features within classes regardless of the class sample size (to alleviate the issues of data scarcity and data density) and simultaneously preserves the semantic correlations between classes using learnable similarity when pushing different classes away from each other (to reduce overlapping). In summary, our IDID-loss provides three advantages: 1) it can simultaneously mitigate all the three issues while DML and CIL losses cannot; 2) it generates more diverse and discriminant feature representations with higher generalization ability, compared with DML losses; and 3) it provides a larger improvement on the classes of data scarcity and density with a smaller sacrifice on easy class accuracy, compared with CIL losses. Experimental results on seven public real-world datasets show that our IDID-loss achieves the best performances in terms of G-mean, F1-score, and accuracy when compared with both state-of-the-art (SOTA) DML and CIL losses. In addition, it gets rid of the time-consuming fine-tuning process over the hyperparameters of loss function.

XAI-FruitNet: An explainable deep model for accurate fruit classification

In agricultural technology, precise fruit classification is essential yet challenging due to inherent interclass similarities and intra-class variabilities among fruit species. Despite their impressive performance, traditional deep learning models suffer from a lack of interpretability, which hampers their transparency and trustworthiness in practical applications. To address these issues, we present XAI-FruitNet, a novel hybrid deep learning architecture designed to enhance feature discrimination by integrating average and max pooling techniques. XAI-FruitNet, an optimized architecture for efficiency evaluated across the Fruits-360, Fruit Recognition, Fruit and Vegetables Image Recognition, and Dry Fruit datasets, consistently achieves over 97 % accuracy, surpassing existing state-of-the-art models and underscoring its remarkable generalization capability. A significant advancement of XAI-FruitNet is its built-in interpretability, which enhances the model's transparency and fosters trust among endusers. Through rigorous experimentation, we demonstrate that XAI-FruitNet advances state-of-the-art fruit classification accuracy and sets a new standard for explainable artificial intelligence (XAI) in agricultural applications. This hybrid approach ensures that stakeholders can rely on the classification outcomes' high performance and comprehensible nature, thereby offering a robust and trustworthy solution for modern agricultural needs.

Intelligent Manufacturing in Wine Barrel Production: Deep Learning-Based Wood Stave Classification

Innovative wood inspection technology is crucial in various industries, especially for determining wood quality by counting rings in each stave, a key factor in wine barrel production. (1) Background: Traditionally, human inspectors visually evaluate staves, compensating for natural variations and characteristics like dirt and saw-induced aberrations. These variations pose significant challenges for automatic inspection systems. Several techniques using classical image processing and deep learning have been developed to detect tree-ring boundaries, but they often struggle with woods exhibiting heterogeneity and texture irregularities. (2) Methods: This study proposes a hybrid approach combining classical computer vision techniques for preprocessing with deep learning algorithms for classification, designed for continuous automated processing. To enhance performance and accuracy, we employ a data augmentation strategy using cropping techniques to address intra-class variability in individual staves. (3) Results: Our approach significantly improves accuracy and reliability in classifying wood with irregular textures and heterogeneity. The use of explainable AI and model calibration offers a deeper understanding of the model’s decision-making process, ensuring robustness and transparency, and setting confidence thresholds for outputs. (4) Conclusions: The proposed system enhances the performance of automatic wood inspection technologies, providing a robust solution for industries requiring precise wood quality assessment, particularly in wine barrel production.

Multi-view visual semantic embedding for cross-modal image–text retrieval

Visual Semantic Embedding (VSE) is a dominant method for cross-modal image–text retrieval. The purpose of VSE is to learn an embedding space where images can be embedded close to the corresponding captions. However, there are large intra-class variations in image–text data. Multiple captions describing the same image may be described from different views, and descriptions of different views are often dissimilar. The VSE method embeds samples from the same class in similar positions, which suppresses intra-class variations and leads to inferior generalization. This paper proposes a Multi-View Visual Semantic Embedding (MV-VSE) framework that learns multiple embeddings for an image, explicitly modeling intra-class variation. To optimize the MV-VSE framework, a multi-view triplet loss is proposed, which jointly optimizes multi-view embeddings while retaining intra-class variation. Recently, large-scale Vision-Language Pre-training (VLP) has become a new paradigm for cross-modal image–text retrieval. To allow our framework to be flexibly applied to the traditional VSE models and VSE-based VLP models, we incorporate the contrastive loss commonly used in VLP and the triplet loss into a unified loss, and further propose a multi-view unified loss. Our framework can be applied plug-and-play to traditional VSE models and VSE-based VLP models without excessively increasing model complexity. Experimental results on the image–text retrieval benchmark datasets demonstrate that applying our framework can boost the retrieval performance of current VSE models. The code is available at https://github.com/AAA-Zheng/MV-VSE.

Universal Fingerprint Generation: Controllable Diffusion Model with Multimodal Conditions.

The utilization of synthetic data for fingerprint recognition has garnered increased attention due to its potential to alleviate privacy concerns surrounding sensitive biometric data. However, current methods for generating fingerprints have limitations in creating impressions of the same finger with useful intra-class variations. To tackle this challenge, we present GenPrint, a framework to produce fingerprint images of various types while maintaining identity and offering humanly understandable control over different appearance factors such as fingerprint class, acquisition type, sensor device, and quality level. Unlike previous fingerprint generation approaches, GenPrint is not confined to replicating style characteristics from the training dataset alone: it enables the generation of novel styles from unseen devices without requiring additional fine-tuning. To accomplish these objectives, we developed GenPrint using latent diffusion models with multimodal conditions (text and image) for consistent generation of style and identity. Our experiments leverage a variety of publicly available datasets for training and evaluation. Results demonstrate the benefits of GenPrint in terms of identity preservation, explainable control, and universality of generated images. Importantly, the GenPrint-generated images yield comparable or even superior accuracy to models trained solely on real data and further enhances performance when augmenting the diversity of existing real fingerprint datasets.

Decomposed Prototype Learning for Few-Shot Scene Graph Generation

Today's scene graph generation (SGG) models typically require abundant manual annotations to learn new predicate types. Therefore, it is difficult to apply them to real-world applications with massive uncommon predicate categories whose annotations are hard to collect. In this paper, we focus on Few-Shot SGG (FSSGG) , which encourages SGG models to be able to quickly transfer previous knowledge and recognize unseen predicates well with only a few examples. However, current methods for FSSGG are hindered by the high intra-class variance of predicate categories in SGG: On one hand, each predicate category commonly has multiple semantic meanings under different contexts. On the other hand, the visual appearance of relation triplets with the same predicate differs greatly under different subject-object compositions. Such great variance of inputs makes it hard to learn generalizable representation for each predicate category with current few-shot learning (FSL) methods. However, we found that this intra-class variance of predicates is highly related to the composed subjects and objects. To model the intra-class variance of predicates with subject-object context, we propose a novel Decomposed Prototype Learning (DPL) model for FSSGG. Specifically, we first construct a decomposable prototype space to capture diverse semantics and visual patterns of subjects and objects for predicates by decomposing them into multiple prototypes. Afterwards, we integrate these prototypes with different weights to generate query-adaptive predicate representation with more reliable semantics for each query sample. We conduct extensive experiments and compare with various baseline methods to show the effectiveness of our method.