Camera-only 3D object detection has emerged as a cost-effective and scalable alternative to LiDAR for autonomous driving, yet existing methods primarily prioritize overall performance while overlooking the severe long-tail imbalance inherent in real-world datasets. In practice, many rare but safety-critical categories such as children, strollers, or emergency vehicles are heavily underrepresented, leading to biased learning and degraded performance. This challenge is further exacerbated by pronounced inter-class ambiguity (e.g., visually similar subclasses) and substantial intra-class diversity (e.g., objects varying widely in appearance, scale, pose, or context), which together hinder reliable long-tail recognition. In this work, we introduce SemLT3D, a Semantic-Guided Expert Distillation framework designed to enrich the representation space for underrepresented classes through semantic priors. SemLT3D consists of: (1) a language-guided mixture-of-experts module that routes 3D queries to specialized experts according to their semantic affinity, enabling the model to better disentangle confusing classes and specialize on tail distributions; and (2) a semantic projection distillation pipeline that aligns 3D queries with CLIP-informed 2D semantics, producing more coherent and discriminative features across diverse visual manifestations. Although motivated by long-tail imbalance, the semantically structured learning in SemLT3D also improves robustness under broader appearance variations and challenging corner cases, offering a principled step toward more reliable camera-only 3D perception.

We investigate how to make small tabular foundation models effective for High-Dimensional, Low-Sample Size (HDLSS) tabular prediction without retraining large backbones. We introduce Graph-guided Ordering with Local Refinement (GO-LR), show its equivalence to weighted Minimum Linear Arrangement, and interpret the practical solver as a TSP-path-style surrogate. We propose GOTabPFN,which builds on GO-LR, and a Neuro-Inspired Subunit Compression (NSC) unit to pool locally adjacent ordered features into meta-features, yielding a compact representation that makes TabPFN-style prediction practical in HDLSS regimes. Across tabular benchmarks, GOTabPFN improves stability and accuracy under tight token budgets.

High-Dimensional Low-Sample Size (HDLSS) tabular domains (e.g., omics) are characterized by n ≪ m, where n = number of samples, and m = number of features. Such domains often exhibit strong local correlation groups, sparse cross-group dependencies, heavy-tailed non-Gaussian marginals, heteroscedastic noise, and structured missingness, making direct density learning in ℝ^m ill-conditioned since n ≪ m. We propose BSTabDiff, a block-subunit generative framework that partitions the m observed features into M latent blocks (M ≪ m) and generates each block via a shared low-dimensional subunit variable, concentrating global dependence learning in the compact block-latent space ℝ^M while decoding to the full feature space with copula-driven dependence, flexible per-feature marginals, and explicit missingness mechanisms. BSTabDiff supports modern deep priors on block latents, including diffusion and normalizing flows, enabling stable synthesis and controllable benchmark generation in the HDLSS regime. Empirically, BSTabDiff produces more realistic and stable high-dimensional synthetic data when compared with unstructured tabular generators on HDLSS data.

High-dimensional tabular data lacks a natural feature order, limiting the applicability of permutation-sensitive deep learning models. We propose DynaTab, a dynamic feature ordering-enabled architecture inspired by neural rewiring. We introduce a lightweight criterion that predicts when feature permutation will benefit a dataset by quantifying its intrinsic complexity. DynaTab dynamically reorders features via a neural rewiring algorithm and processes them through a compact, dynamic order-aware combination of separate learned positional embedding, importance-based gating, and masked attention layers, compatible with any sequence-sensitive backbone. Trained end-to-end with bespoke dynamic feature ordering (DFO) and dispersion losses, DynaTab achieves statistically significant gains, particularly on high-dimensional datasets, where it is benchmarked against 45 state-of-the-art baselines across 36 different real-world tabular datasets. Our results position DynaTab as a compelling new paradigm for high-dimensional tabular deep learning.

Multimodal learning of images and tabular data is often impaired by ineffective representations, resulting in redundancy, dispersion, and generalization problems. To tackle this challenge, we introduce Graph-Enhanced Descriptor Sequencing (GEDS), a structured feature sequencing algorithm grounded in principles from the Column Permutation Problem (CPP). GEDS refines statistical descriptors of the features through similarity graph-based computations, systematically determining an effective feature sequencing. We incorporate GEDS within an order-aware efficient transformer framework, utilizing order-aware memory tokens that explicitly adhere to the derived feature sequencing via a dedicated loss function. Experimental results across multimodal benchmarks demonstrate that iStructTab effectively minimizes feature dispersion, improving predictive performance and robustness, and highlighting the significance of structured feature sequencing in multimodal learning. See code: https://github.com/zadid6pretam/iStructTab

Proper decoding of human emotions based on physiological electroencephalography (EEG) signals significantly contributes to the development of human-computer interface related applications. Current major challenges hindering the recognition performance include the following: 1) the high variance and unknown noise that exist in the EEG recordings; 2) the size of EEG datasets are relatively small given the acquisition effort and annotation cost. It is worthwhile to explore approaches to improve decoding robustness under low data resource scenarios. Previous works utilized data augmentation techniques to tackle this problem using manually designed augmentation operations, leading to sub-optimal performance. In this study, we propose a principled framework to perform dynamic evolution on signal data and improve robustness in the occurrence of unknown corruptions or variances. The framework is formed with bi-level distributionally robust optimization (DRO), and improves robustness by simultaneously optimizing on a family of evolved distributions instead of the single training data distribution. We transform the formed gradient flow system into different types of concrete evolution instantiations based on Langevin dynamics and Hamiltonian dynamics, with tailored divergence measures serving as distance constraint. We performed extensive evaluation of the proposed approach on datasets covering different types of affective states, with model robustness tested on different types of corruptions and adversarial examples. The model outperforms competitive baselines by a significant margin on these challenging emotion recognition benchmarks, especially for low data resource scenarios.

Agents operating in the open-world sense data streams that are non-stationary and temporally correlated, from which they need to quickly learn new knowledge to adapt and be resilient to the changing environment they explore. Current learning-based vision approaches are not designed for agents operating in real-time in the open-world. They are based on stochastic optimizations performed offline with data samples independent and identically distributed according to a stationary probability distribution. Data streams processed by open-world agents have temporal dependencies, and are typically processed with a single-pass. In such conditions, vanilla SGD leads to slow convergence rates and biased estimations, and current approaches make simplifying assumptions, like having prior knowledge of the statistical properties of the stream. We overcome these limitations by introducing an online stochastic optimization approach that does not make any prior assumption and self-adapts to the current time dependency properties of the stream. We demonstrate the validity of the approach on synthetic data as well as on video datasets where we show very promising results for learning self-supervised representations and for classification downstream tasks.

Survival analysis is a crucial statistical technique used to estimate the anticipated duration until a specific event occurs. However, current methods often involve discretizing the time scale and struggle with managing absent features within the data. This becomes especially pertinent since events can transpire at any given point, rendering event analysis a continuous concern. Additionally, the presence of missing attributes within tabular data is widespread. By leveraging recent developments of Transformer and Self-Supervised Learning (SSL), we introduce SSL-SurvFormer. This entails a continuously monotonic Transformer network, empowered by SSL pre-training, that is designed to address the challenges presented by continuous events and absent features in survival prediction. Our proposed continuously monotonic Transformer model facilitates accurate estimation of survival probabilities, thereby bypassing the need for temporal discretization. Additionally, our SSL pre-training strategy incorporates data transformation to adeptly manage missing information. The SSL pre-training encompasses two tasks: mask prediction, which identifies positions of absent features, and reconstruction, which endeavors to recover absent elements based on observed ones. Our empirical evaluations conducted across a variety of datasets, including FLCHAIN, METABRIC, and SUPPORT, consistently highlight the superior performance of SSL-SurvFormer in comparison to existing methods. Additionally, SSL-SurvFormer demonstrates effectiveness in handling missing values, a critical aspect often encountered in real-world datasets.

Adverse drug events represent a key challenge in public health, especially concerning drug safety profiling and drug surveillance. Drug-drug interactions represent one of the most popular types of adverse drug events. Most computational approaches to this problem have used different types of drug-related information utilizing different machine-learning algorithms to predict potential drug interactions. In this work, we focus on genetic information about the drugs, particularly the protein sequence and protein structure of protein targets in drug interaction networks, to predict potential drug interactions. We collected various drug information like drug-drug interaction (DDI), Drug attributes like drug active ingredients, protein targets, protein sequence, protein structure etc. We proposed a similarity-based Neural Network framework called protein sequence-structure similarity network (PS3N) and used this to predict novel DDI's. The drug-drug similarities are computed using different categories of drug information based on multiple similarity metrics. Our method outperforms the state-of-the-art and achieves competitive results. Our performance evaluations on different datasets showed the predictive performance as follows: Precision 91%-98%, Recall 90%-96%, F1 Score 86%-95%, Area Under Curve (AUC) 88%-99%, and Accuracy 86%-95%. Our evaluation demonstrates the effectiveness of PS3N in predicting DDI's, including the clinical significance of some new DDI's discovered by the model.

Using Deep Learning in computer-aided diagnosis systems has been of great interest due to its impressive performance in the general domain and medical domain. However, a notable challenge is the lack of explainability of many advanced models, which poses risks in critical applications such as diagnosing findings in CXR. To address this problem, we propose ItpCtrl-AI, a novel end-to-end interpretable and controllable framework that mirrors the decision-making process of the radiologist. By emulating the eye gaze patterns of radiologists, our framework initially determines the focal areas and assesses the significance of each pixel within those regions. As a result, the model generates an attention heatmap representing radiologists’ attention, which is then used to extract attended visual information to diagnose the findings. By allowing the directional input, our framework is controllable by the user. Furthermore, by displaying the eye gaze heatmap which guides the diagnostic conclusion, the underlying rationale behind the model’s decision is revealed, thereby making it interpretable.In addition to developing an interpretable and controllable framework, our work includes the creation of a dataset, named Diagnosed-Gaze++, which aligns medical findings with eye gaze data. Our extensive experimentation validates the effectiveness of our approach in generating accurate attention heatmaps and diagnoses. The experimental results show that our model not only accurately identifies medical findings but also precisely produces the eye gaze attention of radiologists. The dataset, models, and source code will be made publicly available upon acceptance.

The increased popularity of digital healthcare services has prompted the development of different types of datadriven healthcare APIs on top of electronic health records, offering the convenience of aided diagnostic services without compromising privacy. Defense against the unauthorized extraction of healthcare APIs is important due to: 1) the unauthorized cloned model could serve online as fake healthcare service providers and pose harm to the general public; 2) protected training data containing private electronic health records could be further extracted from the stolen model. It is therefore important to protect the data-driven healthcare APIs from unauthorized clone and extraction. In this work, we propose a principled defense strategy with adaptive teleportation of incoming queries to effectively guard against extraction attacks of healthcare APIs. The proposed mechanism prevents unauthorized copy of model functionality while maintaining the utility of APIs to serve benign queries. The adaptive teleportation operations are generated based on the formulated bi-level optimization target and follows the evolution trajectory depicted by the Wasserstein gradient flows, which effectively push attacking queries to cross decision boundary while constraining the deviation level of benign queries, utilizing the fact that attacker generated pseudo-queries are mostly closer to decision boundaries than normal queries. This provides misleading information on malicious queries while preserving model utility. We performed detailed analysis of the proposed mechanism on three healthcare related prediction tasks including in-hospital mortality, bleed risk and ischemic risk prediction for validation of its effectiveness under different types of attacking scenarios. The proposed mechanism is significantly more effective to suppress the performance of cloned model while maintaining comparable serving utility compared to existing defense approaches.

The multiclass classification of tiny objects of images taken from unmanned aerial vehicles (UAVs) is a challenging task within the field of computer vision research. In order to address this challenge, we propose an efficient patch-based tiny object classification framework. To improve the classification performance and optimize the model efficiency, we present the one-vs-all binary classification-based problem decomposition approach, a z-score-based cluster sampling for selecting the discriminative patches to train the binary classifiers, integration of the decisions from different classifiers at patch level, and finally application of a neighborhood-based image-level decision-fusion to get the image-level label. Using a high-resolution aerial image database, we demonstrate the effectiveness of our approach.

Efficient visual trackers overfit to their training distributions and lack generalization abilities, resulting in them performing well on their respective in-distribution (ID) test sets and not as well on out-of-distribution (OOD) sequences, imposing limitations to their deployment in-the-wild under constrained resources. We introduce Siam-ABC, a highly efficient Siamese tracker that significantly improves tracking performance, even on OOD sequences. SiamABC takes advantage of new architectural designs in the way it bridges the dynamic variability of the target, and of new losses for training. Also, it directly addresses OOD tracking generalization by including a fast backward-free dynamic test-time adaptation method that continuously adapts the model according to the dynamic visual changes of the target. Our extensive experiments suggest that Siam-ABC shows remarkable performance gains in OOD sets while maintaining accurate performance on the ID benchmarks. SiamABC outperforms MixFormerV2-S by 7.6% on the OOD AVisT benchmark while being 3x faster (100 FPS) on a CPU. Our code and models are available at https://wvuvl.github.io/SiamABC/.

Effective analysis of tabular data still poses a significant problem in deep learning, mainly because features in tabular datasets are often heterogeneous and have different levels of relevance. This work introduces TabSeq, a novel framework for the sequential ordering of features, addressing the vital necessity to optimize the learning process. Features are only sometimes equally informative, and for certain deep learning models, their random arrangement can hinder the model’s learning capacity. Finding the optimum sequence order for such features could improve the deep learning models’ learning process. The novel feature ordering technique, which we provide in this work, is based on clustering and incorporates both local ordering and global ordering. It is designed to be used with a multi-head attention mechanism in a denoising autoencoder network. Our framework uses clustering to align comparable features and improve data organization. Multi-head attention focuses on essential characteristics, whereas denoising autoencoder highlights important aspects by rebuilding from distorted inputs. This method improves the capability to learn from tabular data while lowering redundancy. Our research demonstrating improved performance through appropriate feature sequence rearrangement utilizing raw antibody microarray and two other real-world biomedical datasets validates the impact of feature ordering. These results demonstrate that feature ordering can be a viable approach to improved deep learning of tabular data.

Despite recent significant progress, Multi-Object Tracking (MOT) faces limitations such as reliance on prior knowledge and pre-defined categories and struggles with unseen objects. To address these issues, Generic Multiple Object Tracking (GMOT) has emerged as an alternative approach, requiring less prior information. However, current GMOT methods often rely on initial bounding boxes and struggle to handle variations in factors such as viewpoint, lighting, occlusion, and scale, among others. Our contributions commence with the introduction of the Referring GMOT dataset a collection of videos, each accompanied by detailed textual descriptions of their attributes. Subsequently, we propose Z - GMOT, a cutting-edge tracking solution capable of tracking objects from never-seen categories without the need of initial bounding boxes or predefined categories. Within our Z - GMOT framework, we introduce two novel components: (i) iGLIP, an improved Grounded language-image pretraining, for accurately detecting unseen objects with specific characteristics. (ii) MA - SORT, a novel object association approach that adeptly integrates motion and appearance-based matching strategies to tackle the complex task of tracking objects with high similarity. Our contributions are benchmarked through extensive experiments conducted on the Referring GMOT dataset for GMOT task. Additionally, to assess the generalizability of the proposed Z - GMOT, we conduct ablation studies on the DanceTrack and MOT20 datasets for the MOT task. Our dataset, code, and models are released at: https://fsoft-aic.github.io/Z-GMOT.

Large variance exists in Electroencephalogram (EEG) signals with its pattern differing significantly across subjects. It is a challenging problem to perform online sequential decoding of EEG signals across different subjects, where a sequence of subjects arrive in temporal order and no signal data is jointly available beforehand. The challenges include the following two aspects: 1) the knowledge learned from previous subjects doesn’t readily fit to future subjects, and fast adaptation is needed in the process; 2) the EEG classifier could drastically erase information of learnt subjects as learning progresses, namely catastrophic forgetting. Most existing EEG decoding explorations use sizable data for pre-training purposes, and to the best of our knowledge we are the first to tackle this challenging online sequential decoding setting. In this work, we propose a unified bi-level meta-learning framework that enables the EEG decoder to simultaneously perform fast adaptation on future subjects and retain knowledge of previous subjects. In addition, we extend to the more general subject-agnostic scenario and propose a subject shift detection algorithm for situations that subject identity and the occurrence of subject shifts are unknown. We conducted experiments on three public EEG datasets for both subject-aware and subject-agnostic scenarios. The proposed method demonstrates its effectiveness in most of the ablation settings, e.g. an improvement of 5.73% for forgetting mitigation and 3.50% for forward adaptation on SEED dataset for subject agnostic scenarios.

Precise 3D environmental mapping is pivotal in robotics. Existing methods often rely on predefined concepts during training or are time-intensive when generating semantic maps. This paper presents Open-Fusion, a groundbreaking approach for real-time open-vocabulary 3D mapping and queryable scene representation using RGB-D data. Open-Fusion harnesses the power of a pre-trained vision-language foundation model (VLFM) for open-set semantic comprehension and employs the Truncated Signed Distance Function (TSDF) for swift 3D scene reconstruction. By leveraging the VLFM, we extract region-based embeddings and their associated confidence maps. These are then integrated with 3D knowledge from TSDF using an enhanced Hungarian-based feature-matching mechanism. Notably, Open-Fusion delivers outstanding annotation-free 3D segmentation for open-vocabulary without necessitating additional 3D training. Benchmark tests on the ScanNet dataset against leading zero-shot methods highlight Open-Fusion's superiority. Furthermore, it seamlessly combines the strengths of region-based VLFM and TSDF, facilitating real-time 3D scene comprehension that includes object concepts and open-world semantics. We encourage the readers to view the demos on our project page: https://uark-aicv.github.io/OpenFusion

Developing an interpretable system for generating reports in chest X-ray (CXR) analysis is becoming increasingly crucial in Computer-aided Diagnosis (CAD) systems, enabling radiologists to comprehend the decisions made by these systems. Despite the growth of diverse datasets and methods focusing on report generation, there remains a notable gap in how closely these models's generated reports align with the interpretations of real radiologists. In this study, we tackle this challenge by initially introducing Fine-Grained CXR (FG-CXR) dataset, which provides fine-grained paired information between the captions generated by radiologists and the corresponding gaze attention heatmaps for each anatomy. Unlike existing datasets that include a raw sequence of gaze alongside a report, with significant misalignment between gaze location and report content, our FG-CXR dataset offers a more grained alignment between gaze attention and diagnosis transcript. Furthermore, our analysis reveals that simply applying black-box image captioning methods to generate reports cannot adequately explain which information in CXR is utilized and how long needs to attend to accurately generate reports. Consequently, we propose a novel explainable radiologist's attention generator network (Gen-XAI) that mimics the diagnosis process of radiologists, explicitly constraining its output to closely align with both radiologist's gaze attention and transcript. Finally, we perform extensive experiments to illustrate the effectiveness of our method. Our datasets and checkpoint is available at https://github.com/UARK-AICV/FG-CXR.

Developing multi-label classification models under significant class imbalance, and when annotating data requires expert-level knowledge remains a major challenge. Additionally, interdependency and correlation among labels are common in multi-label problems. In this work, we introduce a novel framework to address these challenges. Our approach extracts robust discriminative features from unlabeled data through self-supervised contrastive learning and uses an adaptive data augmentation mechanism (ACBA) to balance the dataset. Independent binary classifiers are trained for each class, using a new custom Focal Weighted Cross-Entropy (FWCE) loss function to focus on hard-to-classify examples. A correlation learning module then refines predictions by integrating statistical and domain-specific knowledge. Finally, a meta-learner, employing a Gated Recurrent Unit (GRU) and multi-head attention, identifies complex relationships between classes, even for those that rarely occur together. We used the detection of thoracic diseases using chest X-rays, a domain with a major class imbalance and highly associated labels, to validate our approach. Our findings demonstrate the potential of our method to apply to other medical and non-medical imaging scenarios with similar multi-label classification problems.