Generalized detection of diffusion-generated forgeries across unseen generators is essential for maintaining visual authenticity in the era of generative media. Recent studies incorporate pretrained diffusion models into detection frameworks, primarily capturing forgery cues via image-level reconstruction through the complete diffusion-denoising process. However, such schemes lack an explicit separation between the semantic priors encoded in diffusion representations and the generator-specific noise trajectories introduced during synthesis, causing detectors to overfit to artifacts from specific training generators. To address this limitation, we propose a two-stage diffusion-guided framework that exploits latent-space priors within a pretrained denoising backbone and decouples the detection process into authentic manifold modeling and forgery discrimination stages. In the authentic manifold modeling phase, stable real-image representations are distilled from diffusion features using only authentic data to establish a generator-agnostic semantic space. During forgery discrimination, forged samples are incorporated and multi-layer residual modeling is designed to capture representation inconsistencies relative to the distilled real-image manifold. Extensive experiments conducted on multiple diffusion forgery benchmarks demonstrate that the proposed approach achieves consistent cross-generator generalization performance.

To address the lack of generalized feature representation in cross-domain electroencephalogram emotion recognition, this letter proposes a multi-source domain adaptation model that integrates prototype-based class-level constraints into a mutual information disentanglement mechanism. The model mainly consists of a mutual information disentanglement module that separates domain-related and domain-unrelated features by minimizing mutual information, and a prototype classification module that enhances intra-class compactness and semantic consistency across domains. To align multiple source domains with the target domain, Central Moment Discrepancy is minimized in the prototype space. Unlike prior multi-branch MSDA models, our method uses a unified feature extractor, enhancing scalability. Experiments show our model outperforms existing domain adaptation models and achieves high accuracy and stability.

In machine learning, the relatedness across multiple tasks is usually complex and entangled. Due to dataset bias, the relatedness among tasks might be distorted and mislead the training of the models with solid learning ability, such as the multi-task neural networks. In this paper, we propose the idea of Relatedness Refinement Multi-Task Learning (RRMTDL) by introducing adversarial learning in the multi-task deep neural network to tackle the problem. The RRMTDL deep learning model restrains the misleading relatedness task by adversarial training and extracts information sharing across tasks with valuable relatedness. With RRMTDL, multi-task deep learning can enhance the task-specific representation for the major tasks by excluding the misleading relatedness. We design tests with various combinations of task-relatedness to validate the proposed model. Experimental results show that the RRMTDL model can effectively refine the task relatedness and prominently outperform other multi-task deep learning models in datasets with entangled task labels.

Pedestrian attribute recognition aims to predict human attributes in surveillance video. Since the correlations are different among attribute groups, mining the correlation variation between attributes is an effective way to improve the performance of pedestrian attribute recognition. In this paper, we propose a Feature Enhanced Multi-scale and Graph Convolutional Compound Network named FEM-GCNet to learn the enhanced correlations between attributes. The FEM-GCNet contains an enhanced feature extractor and a graph convolutional network. The enhanced feature extractor learns the enhanced features of attributes, improving the feature representation. The graph convolutional network exploits attribute correlation rules to mine the correlation features between attributes, which are embedded into the enhanced features to obtain the enhanced attribute correlations. Finally, a multi-label loss is used to train our model to identify pedestrian attributes. Experiments on RAPv1 and RAPv2 datasets demonstrate the robustness of the enhanced attribute correlations and the better performance of the FEM-GCNet.

Sign language recognition (SLR) has become increasingly popular in recent years in computer vision. It is essential to extract discriminative spatiotemporal features to model the spatial and temporal evolutions of different signs. Also, local gesture and facial expression representations contribute to distinguishing signs with similar motion patterns but different meanings. In this paper, we propose a multi-modal sign language recognition framework, in the RGB representation model, we design the adaptive spatiotemporal attention modules to fulfill the visual cue definition in signing videos, and the adapter is designed for constructing the auxiliary task, which jointly learning with the SLR task to enhances the performance of the model. Given a signing video, a spatiotemporal attention-based Pseudo-3D Residual Networks (STA P3D ResNet) is used to learn spatiotemporal features mainly from the areas of interest and the key frames. After feature extraction, the attention-based Bidirectional Long Short-Term Memory Networks (Att-BLSTM) is utilized to select the significant motions. Meanwhile, from the skeletal data, we can obtain the texture image by color encoding and construct the spatial relation features, which are high level representations of human posture. The learnt skeleton-based features from skeletal data fused with the attention-aware video features to further provide more informative spatiotemporal information for SLR. Experiments are carried out on two large scale sign language datasets. And the experimental results demonstrate the effectiveness of our proposed method.

Interactive face retrieval aims at finding target subjects in face databases through human and machine interaction, which involves user feedback based on human perception and machine similarity measure in feature spaces. In this article, we propose an attribute prototype learning method to tackle the semantic gap between human and machine in face perception for fast interactive face retrieval. We reformulate the theoretical explanation of the interactive retrieval model and develop the algorithm of the heuristic solution of the model. Each module of the prototype model is learned with a set of identity-related facial attributes. The outputs of the prototype modules form the semantic representation. To adapt the prototype models across different databases, we propose a transfer selection algorithm based on the coherence measurements in interactive face retrieval. Coherence analysis proves that the proposed attribute prototype representation can effectively narrow down the semantic gap even in the case of cross-database transfer learning. The prototype representation can effectively reduce the feature dimension in the retrieval process. Real user retrieval with the Bayesian relevance feedback model shows that attribute prototype space is superior to low-level feature space and proves that interactive retrieval with attribute prototype representation can converge fast in large face databases.

In the treatment of epilepsy with intracranial electroencephalogram(iEEG), the recognition accuracy is low, and it is difficult to find the correlation between channels because of the large amount of channel numbers and time series data. In order to solve these problems, we propose a novel EEG feature pre-presentation method for seizure detection based on the Log Mel-Filterbank energy feature. We propose to adapt the Mel-Filterbank energy to EEG features with logrithm transform in the frequency domain. Meanwhile, we also propose the sequential forward channel selection(SFCS) algorithm to incorporate channel correlation and balance the computing consumption. Experiments show that our proposed method have significant contributions in channel selection and feature representation to the problem of EEG signal analysis. The average results of experiments judged by the mean area under the ROC curve (AUC) of the probability reach 99.13%.

In this article, we explore the coherence of face perception between human and machine in the scenario of interactive face retrieval. In the part of human perception, we collect user feedback to the stimuli of a target face and groups of displayed candidate face images in a face database with a large number of subjects. In the part of machine vision, we compare the benchmark features and general metrics to measure face similarity. We propose a series of coherence measurements to evaluate the statistic characteristic of human and machine face perception. We discover that despite the unfamiliarity of users to most faces in the database, the coherence between human and machine remains in a stable level across multiple variations in metrics, features, size of databases, and demographics. The simulation experiments with the coherence distributions demonstrate that the embedded information is valuable to speed up interactive retrieval. The comparisons over multiple parameter settings provide feasible instructions in designing the interactive face retrieval system with more consideration of human factors.

In the attribute recognition area, attributes that are unrelated in the real world may have a high co-occurrence rate in a dataset due to the dataset bias, which forms a misleading relatedness. A neural network, especially a multi-task neural network, trained on this dataset would learn this relatedness, and be misled when it is used in practice. In this paper, we propose Share-and-Compete Multi-Task deep learning (SCMTL) model to handle this problem. This model uses adversarial training methods to enhance competition between unrelated attributes while keeping sharing between related attributes, making the task-specific layer of the multi-task model to be more specific and thus rule out the misleading relatedness between the unrelated attributes. Experiments performed on elaborately designed datasets show that the proposed model outperforms the single task neural network and the traditional multi-task neural network in the situation mentioned above.

Automatic emotion recognition based on electroencephalo-graphic (EEG) signals has received increasing attention in recent years. The Deep Residual Networks (ResNets) can solve vanishing gradient problem and exploding gradient problem well in computer vision and can learn more profound semantic information. And for traditional methods, frequency features often play important role in signal processing area. Thus, in this paper, we use the pre-trained ResNets to extract deep semantic information and the linear-frequency cepstral coefficients (LFCC) as features from raw EEG signals. Then the two features are fused to improve the emotion classification performance of our approach. Moreover, several classifiers are used for our fused features to evaluate the performance and it shows that the proposed approach is effective for emotion classification. We find that the best performance is achieved when use k-nearst neighbor (KNN) as classifier, and we provide a detailed discussion for the reason.

The application cost machine learning methods often rely on the availability of large-scale data collection and annotation, especially in the cases of cross-domain learning. One way to circumvent this cost is constructing models to synthesize data and provide automatic annotation. Although these models are attractive, they often can not be generalized from synthetic images to real-world images. Therefore, domain adaptive algorithm is needed to improve these models, so that they can be applied successfully. In this paper, we propose a novel unsupervised domain adaptive framework codenamed D-DANN inspired by the theory of adversarial learning. We apply the discriminator to diverse the features extracted from dual branch CNN. We can obtain more sufficient shared representation across domains by the proposed dual feature extractors. The framework can be easily adapt to most popular CNN models to improve the representation power. We implement the D-DANN with several popular CNN models including LeNet, AlexNet and so on. Using these D-DANN enhanced neural networks, we conduct extensive experiments on several pairs of domain adaptive validation datasets. The results show that our approach can efficiently enhance domain adaptive capability of general CNN models for unlabeled data.

In this paper, we propose an improved attribute hypergraph learning framework and adapt it for metric learning. Under the attribute hypergraph, each image is abstracted as a vertex and is contained in some hyperedges, each of which represents an attribute. The learned attribute hypergraph mines the correlation among multiple facial attributes and serves to reform the topology of the image similarity relationship in a database defined by the distance metric. By combining the merits of attribute and hypergraph, the reformed distance metric from the attribute hypergraph learning framework is able to capture the intrinsic information of the data set. Extensive experimental results on LFW data set confirm the effectiveness of the improved attribute hypergraph learning framework.

Multi-task learning and deep convolutional neural network (CNN) have been successfully used in various fields. This paper considers the integration of CNN and multi-task learning in a novel way to further improve the performance of multiple related tasks. Existing multi-task CNN models usually empirically combine different tasks into a group which is then trained jointly with a strong assumption of model commonality. Furthermore, traditional approaches usually only consider small number of tasks with rigid structure, which is not suitable for large-scale applications. In light of this, we propose a dynamic multi-task CNN model to handle these problems. The proposed model directly learns the task relations from data instead of subjective task grouping. Due to its flexible structure, it supports task-wise incremental training, which is useful for efficient training of massive tasks. Specifically, we add a new task transfer connection (TTC) between the layers of each task. The learned TTC is able to reflect the correlation among different tasks guiding the model dynamically adjusting the multiplexing of the information among different tasks. With the help of TTC, multiple related tasks can further boost the whole performance for each other. Experiments demonstrate that the proposed dynamic multi-task CNN model outperforms traditional approaches.

Image Quality Assessment (IQA) aims at automatically predicting the perceptual quality of targets with low computation complexity and high precision. However, it is usually very hard to combine all these merits into one algorithm. In this paper, we propose simple yet efficient facial image quality assessment algorithm - Reduced-Reference Automatic Ranking (RRAR) for face recognition. The RRAR contains a quality control stage and quality ranking stage based on modified structural similarity - Reduced-Reference of SSIM as the reduced reference IQA module. Experimental results show that the proposed algorithm increases the precision of face recognition with low memory consumption and computation complexity and works exceptionally well with face images captured under uncontrolled environment.

Local Binary Patterns (LBP) feature is one of the most popular representation schemes for face recognition. The four factors deciding its effect are the blocking number, image resolution, the sampling radius and sampling density of LBP operator. Numerous previous researches have taken various groups of value of these factors based on experimental comparisons. However, which factor among them contributes the most? Numerous revisions are made to the LBP operators for it is believed that the LBP coding is the most essential factor. Is it true? In this paper, with the very simple and classical Multivariate Analysis of Variance (MANOVA), we discover that the blocking number contributes the most; though all four factors have significant effect for recognition rate. In addition, with the same analysis, we disclose the detailed effect of each factor and their interactions to the precision of LBP features.

Recently, keypoint descriptors such as Scale Invariant Feature Transform (SIFT) have been proved promising in similarity retrieval of images, which adopts matching score as similarity. However, the matching score is easy to be decreased once there are little variances between image details, and hence lead to low retrieval performance. In this paper, we propose a novel retrieval approach that improves the matching score with reduced time of matching by Kernel-based Fuzzy C-Means clustering (KFCM), which proves to be a better trade-off between matching and retrieval precision. Experiments conducted on three representative image databases show that our retrieval approach is surprisingly effective, outperforming the SIFT based method, not only in object-based image retrieval but also for searching scenes with similar semantic.