Large-scale tuberculosis (TB) screening is limited by the high cost and operational complexity of traditional diagnostics, creating a need for artificial-intelligence solutions. We propose DeepGB-TB, a non-invasive system that instantly assigns TB risk scores using only cough audio and basic demographic data. The model couples a lightweight one-dimensional convolutional neural network for audio processing with a gradient-boosted decision tree for tabular features. Its principal innovation is a Cross-Modal Bidirectional Cross-Attention module (CM-BCA) that iteratively exchanges salient cues between modalities, emulating the way clinicians integrate symptoms and risk factors. To meet the clinical priority of minimizing missed cases, we design a Tuberculosis Risk-Balanced Loss (TRBL) that places stronger penalties on false-negative predictions, thereby reducing high-risk misclassifications. DeepGB-TB is evaluated on a diverse dataset of 1,105 patients collected across seven countries, achieving an AUROC of 0.903 and an F1-score of 0.851, representing a new state of the art. Its computational efficiency enables real-time, offline inference directly on common mobile devices, making it ideal for low-resource settings. Importantly, the system produces clinically validated explanations that promote trust and adoption by frontline health workers. By coupling AI innovation with public-health requirements for speed, affordability, and reliability, DeepGB-TB offers a tool for advancing global TB control.

Sign language is the primary communication mode for 72 million hearing-impaired individuals worldwide, necessitating effective bidirectional Sign Language Production and Sign Language Translation systems. However, functional bidirectional systems require a unified linguistic environment, hindered by the lack of suitable unified datasets, particularly those providing the necessary pose information for accurate Sign Language Production (SLP) evaluation. Concurrently, current SLP evaluation methods like back-translation ignore pose accuracy, and high-quality coordinated generation remains challenging. To create this crucial environment and overcome these challenges, we introduce CNText2Sign and CNSign, which together constitute the first unified dataset aimed at supporting bidirectional accessibility systems for Chinese sign language; CNText2Sign provides 15,000 natural language-to-sign mappings and standardized skeletal keypoints for 8,643 vocabulary items supporting pose assessment. Building upon this foundation, we propose the AuraLLM model, which leverages a decoupled architecture with CNText2Sign's pose data for novel direct gesture accuracy assessment. The model employs retrieval augmentation and Cascading Vocabulary Resolution to handle semantic mapping and out-of-vocabulary words, and achieves all-scenario production with controllable coordination of gestures and facial expressions via pose-conditioned video synthesis. Concurrently, our Sign Language Translation model SignMST-C employs targeted self-supervised pretraining for dynamic feature capture, achieving new SOTA results on PHOENIX2014-T with BLEU-4 scores up to 32.08. AuraLLM establishes a strong performance baseline on CNText2Sign with a BLEU-4 score of 50.41 under direct evaluation.

Early detection of retinal lesions is critical for preventing vision loss. While supervised learning has shown promise, existing methods often rely on extensive labeled data, which is costly and difficult to obtain in medical applications. Unsupervised anomaly detection provides an attractive alternative by requiring only healthy retinal images and no abnormal annotations. However, current methods face significant challenges in modeling the complex structures of normal retinal anatomy, learning discriminative features for detecting subtle lesions, and capturing multi-scale features to handle anomalies of varying sizes – highlighting the need for holistic feature modeling that comprehensively represents both retinal anatomy and pathology. To address these challenges, we propose HolistAno, a novel unsupervised anomaly detection framework with holistic retinal modeling. HolistAno adopts a two-stage network architecture, incorporating a novel anomaly generator and a Balanced Mamba Scale Fusion (BMSF) module to effectively learn comprehensive retinal feature representations. This enables accurate detection of subtle lesions, diverse lesion types, and anomalies across multiple scales. Extensive experiments on five benchmark datasets demonstrate that HolistAno achieves state-of-the-art performance in both anomaly classification and localization tasks, with superior generalization and robustness across multiple datasets and cross-dataset scenarios compared to existing methods.

The goal of continual learning (CL) is to acquire new knowledge while retaining previously learned information. CNN-based and prompt-based CL methods have achieved remarkable progress in recent years. However, most prior work has primarily focused on reducing forgetting from the perspective of the model itself. In this paper, we investigate CL from the perspective of decision boundaries, analyzing the impact of instance-level feature overlap. To address this issue, we propose a generic Decoupled Gaussian Softmax Classifier that enhances class discriminability during CL process. Specifically, we decouple the features extracted by the backbone into multiple Gaussian distributions, which are directly fused into the feature space through weighted integration. A regularization term is introduced to penalize the overlap of similar features, while an adaptive decision boundary is assigned to each class to encourage inter-class separation and intra-class compactness. Experiments on 4 widely used continual learning datasets and 12 CL scenarios show that our method has good plug-and-play capability. It improves the average accuracy by 1%–2.63% over the baseline models, while effectively reducing both the forgetting rate and the Expected Calibration Error. Our code is available at: https://anonymous.4open.science/r/DGSC-main-310D .

This paper addresses the fundamental problem of data explosion encountered in the Deep learning-based Direction of Arrival (DOA) estimation, which is caused by the fact that the amount of training data grows exponentially with the number of sources. The data explosion enforces harsh requirements on the computational resources to train the deep-learning (DL) model, even making the training of the DL model mission impossible. To address this, we analyze the similarity between the received array signals in the cases of different sources, which serves as the basis for the proposed training framework. We then give the explicit formula of all possible angle combinations, demonstrating the training data explosion issue. Afterwards, we propose the training framework for progressively fine-tuning the model as the number of sources increases. By reusing a pre-trained model for$k-1$sources and fine-tuning it with a small portion of the data of$k$sources, the method significantly reduces the training overhead. The core of this method lies in utilizing the signal features already learned by the model trained with fewer sources and adapting it to the higher-dimensional source scenario through fine-tuning, thus avoiding the data redundancy associated with training from scratch. Numerical results show the model using the proposed training framework with only 1% of the data achieves similar performance as the fully trained model with 100% of the data.

The goal of Class Incremental Learning (CIL) is to continuously learn new classes while preventing forgetting of old ones. Most previous works focused on reducing catastrophic forgetting from model’s perspective. However, the model is not the only factor contributing to forgetting. In this paper, we take the perspective of class instances and find that fine-grained class increments can lead to feature overlap between classes, further reducing instance margins. We call this interesting phenomenon as Fine-grained class confusion effect in CIL. Since preserving instance margins is crucial for resisting forgetting, it is beneficial to maintain the margin amount as much as possible. To achieve this, we propose a general Gaussian decoupling classifier to enhance the discriminability of similar classes during incremental learning. Specifically, we decouple the features of different classes extracted by the backbone network into multiple independent Gaussian distributions. By directly integrating them into the features with weighted fusion, we introduce a regularization penalty that encourages minimizing the overlap of similar features, thus increasing the feature distance between classes. Extensive experiments show that our method effectively improves class separation and better preserves instance margins, ultimately alleviating forgetting. The improved model achieves better performance on CUB-200 and CARS-196.

Segment-tree (ST) based cost aggregation algorithm for stereo matching successfully integrates the information of segmentation with non-local cost aggregation framework. The tree structure which is generated by the segmentation strategy directly determines the final results for this kind of algorithms. However, the original strategy performs unreasonable due to its coarse performance and ignores to meet the disparity consistency assumption. To improve these weaknesses we propose a novel segmentation algorithm for constructing a more faithful ST with enhanced segmentation advantage according to a robust initial over-segmentation. Then we implement non-local cost aggregation framework on this new ST structure and obtain improved disparity maps. Performance evaluations on all 31 Middlebury stereo pairs show that the proposed algorithm outperforms than other five state-of-the-art aggregated based algorithms and also keeps time efficiency.

The minimum spanning tree (MST) based non-local cost aggregation algorithm performs well in accuracy and time efficiency. However, it can still be improved in two aspects. First, we propose a logarithmic transformation on matching cost function to improve the matching efficiency in texture less regions. The textureless neighbors can provide effective contributions in cost aggregation by the proposed monotone increasing function. Hence the algorithm can distinguish different pixels in textureless regions. Second, MST algorithm only utilizes color information in weight function while aggregating, which leads 3D cues missing. We introduce depth weight computed from the original MST algorithm into an edge weight function. With the proposed color-depth weight, we further iteratively rebuild the tree and obtain enhanced disparity map. Performance evaluations on 19 Middlebury stereo pairs and Microsoft stereo videos show that the proposed algorithm outperforms than other five state-of-the-art cost aggregation algorithms.

Obfuscated malware has become popular because of pure benefits brought by obfuscation: low cost and readily availability of obfuscation tools accompanied with good result of evading signature based anti-virus detection as well as prevention of reverse engineer from understanding malwares' true nature. Regardless obfuscation methods, a malware must deobfuscate its core code back to clear executable machine code so that malicious portion will be executed. Thus, to analyze the obfuscation pattern before unpacking provide a chance for us to prevent malware from further execution. In this paper, we propose a heuristic detection approach that targets obfuscated Windows binary files being loaded into memory - prior to execution. We perform a series of static check on binary file's PE structure for common traces of a packer or obfuscation, and gauge a binary's maliciousness with a simple risk rating mechanism. As a result, a newly created process, if flagged as possibly malicious by the static screening, will be prevented from further execution. This paper explores the foundation of this research, as well as the testing methodology and current results.