• Propose CenterIR, a cluster-aware regularizer for imbalanced EEG regression • Enhance depression severity estimation from resting-state EEG in older adults • Demonstrate stable performance gains over state-of-the-art baselines • Validate robustness through λ -sensitivity and loss ablation analyses Electroencephalography (EEG)-based mental health assessment has gained increasing attention as a non-invasive tool for quantifying depression severity in older adults. However, regression models for continuous severity prediction remain limited, particularly under imbalanced data distributions. This study presents a deep learning framework that integrates convolutional neural networks and bidirectional long short-term memory modules with a novel CenterIR loss to enhance regression performance. Resting-state EEG was recorded from 104 older adults under eyes-open (EO) and eyes-closed (EC) conditions. The proposed model outperformed baseline and state-of-the-art approaches, achieving an MSE of 0.151, MAE of 0.231, and R² of 0.990 in EO, and 0.270, 0.291, and 0.983, respectively, in EC. Paired t-test results indicated significantly better performance in the EO condition, highlighting the potential of EO resting EEG as a reliable neural marker for depression severity. Overall, our framework demonstrates effective imbalanced regression modeling for EEG-based depression assessment in older adults and provides insights for clinical application in mental health monitoring.

We propose a novel knowledge distillation approach, CustomKD, that effectively leverages large vision foundation models (LVFMs) to enhance the performance of edge models (e.g., MobileNetV3). Despite recent advancements in LVFMs, such as DINOv2 and CLIP, their potential in knowledge distillation for enhancing edge models remains underexplored. While knowledge distillation is a promising approach for improving the performance of edge models, the discrepancy in model capacities and heterogeneous architectures between LVFMs and edge models poses a significant challenge. Our observation indicates that although utilizing larger backbones (e.g., ViT-S to ViT-L) in teacher models improves their downstream task performances, the knowledge distillation from the large teacher models fails to bring as much performance gain for student models as for teacher models due to the large model discrepancy. Our simple yet effective CustomKD customizes the well-generalized features inherent in LVFMs to a given student model in order to reduce model discrepancies. Specifically, beyond providing well-generalized original knowledge from teachers, CustomKD aligns the features of teachers to those of students, making it easy for students to understand and overcome the large model discrepancy overall. CustomKD significantly improves the performances of edge models in scenarios with unlabeled data such as unsupervised domain adaptation (e.g., OfficeHome and DomainNet) and semi-supervised learning (e.g., CIFAR-100 and ImageNet), achieving the new state-of-the-art performances.

Despite their consistent performance improvements, cross-modal retrieval models (e.g., CLIP) show degraded performances with retrieving keys composed of fused image-text modality (e.g., Wikipedia pages with both images and text). To address this critical challenge, multimodal retrieval has been recently explored to develop a unified single retrieval model capable of retrieving keys across diverse modality combinations. A common approach involves constructing new composed sets of image-text triplets (e.g., retrieving a pair of image and text given a query image). However, such an approach requires careful curation to ensure the dataset quality and fails to generalize to unseen modality combinations. To overcome these limitations, this paper proposes Generalized Contrastive Learning (GCL), a novel loss formulation that improves multimodal retrieval performance without the burdensome need for new dataset curation. Specifically, GCL operates by enforcing contrastive learning across all modalities within a mini-batch, utilizing existing image-caption paired datasets to learn a unified representation space. We demonstrate the effectiveness of GCL by showing consistent performance improvements on off-the-shelf multimodal retrieval models (e.g., VISTA, CLIP, and TinyCLIP) using the M-BEIR, MMEB, and CoVR benchmarks.

In image classification, debiasing aims to train a classifier to be less susceptible to dataset bias, the strong correlation between peripheral attributes of data samples and a target class. For example, even if the frog class in the dataset mainly consists of frog images with a swamp background (i.e., bias aligned samples), a debiased classifier should be able to correctly classify a frog at a beach (i.e., bias conflicting samples). Recent debiasing approaches commonly use two components for debiasing, a biased model fB and a debiased model fD. fB is trained to focus on bias aligned samples (i.e., overfitted to the bias) while fD is mainly trained with bias conflicting samples by concentrating on samples which fB fails to learn, leading fD to be less susceptible to the dataset bias. While the state of the art debiasing techniques have aimed to better train fD, we focus on training fB, an overlooked component until now. Our empirical analysis reveals that removing the bias conflicting samples from the training set for fB is important for improving the debiasing performance of fD. This is due to the fact that the bias conflicting samples work as noisy samples for amplifying the bias for fB since those samples do not include the bias attribute. To this end, we propose a simple yet effective data sample selection method which removes the bias conflicting samples to construct a bias amplified dataset for training fB. Our data sample selection method can be directly applied to existing reweighting based debiasing approaches, obtaining consistent performance boost and achieving the state of the art performance on both synthetic and real-world datasets.

This paper presents a simple yet effective approach that improves continual test-time adaptation (TTA) in a memory-efficient manner. TTA may primarily be conducted on edge devices with limited memory, so reducing memory is crucial but has been overlooked in previous TTA studies. In addition, long-term adaptation often leads to catastrophic forgetting and error accumulation, which hinders applying TTA in real-world deployments. Our approach consists of two components to address these issues. First, we present lightweight meta networks that can adapt the frozen original networks to the target domain. This novel architecture minimizes memory consumption by decreasing the size of intermediate activations required for backpropagation. Second, our novel self-distilled regularization controls the output of the meta networks not to deviate significantly from the output of the frozen original networks, thereby preserving well-trained knowledge from the source domain. Without additional memory, this regularization prevents error accumulation and catastrophic forgetting, resulting in stable performance even in long-term test-time adaptation. We demonstrate that our simple yet effective strategy outperforms other state-of-the-art methods on various benchmarks for image classification and semantic segmentation tasks. Notably, our proposed method with ResNet-50 and WideResNet-40 takes 86% and 80% less memory than the recent state-of-the-art method, CoTTA.

Despite recent advancements in deep learning, deep neural networks continue to suffer from performance degradation when applied to new data that differs from training data. Test-time adaptation (TTA) aims to address this challenge by adapting a model to unlabeled data at test time. TTA can be applied to pretrained networks without modifying their training procedures, enabling them to utilize a well-formed source distribution for adaptation. One possible approach is to align the representation space of test samples to the source distribution (i.e., feature alignment). However, performing feature alignment in TTA is especially challenging in that access to labeled source data is restricted during adaptation. That is, a model does not have a chance to learn test data in a class-discriminative manner, which was feasible in other adaptation tasks (e.g., unsupervised domain adaptation) via supervised losses on the source data. Based on this observation, we propose a simple yet effective feature alignment loss, termed as Class-Aware Feature Alignment (CAFA), which simultaneously 1) encourages a model to learn target representations in a class-discriminative manner and 2) effectively mitigates the distribution shifts at test time. Our method does not require any hyper-parameters or additional losses, which are required in previous approaches. We conduct extensive experiments on 6 different datasets and show our proposed method consistently outperforms existing baselines.

Test-time adaptation (TTA) methods, which generally rely on the model’s predictions (e.g., entropy minimization) to adapt the source pretrained model to the unlabeled target domain, suffer from noisy signals originating from 1) incorrect or 2) open-set predictions. Long-term stable adaptation is hampered by such noisy signals, so training models without such error accumulation is crucial for practical TTA. To address these issues, including open-set TTA, we propose a simple yet effective sample selection method inspired by the following crucial empirical finding. While entropy minimization compels the model to increase the probability of its predicted label (i.e., confidence values), we found that noisy samples rather show decreased confidence values. To be more specific, entropy minimization attempts to raise the confidence values of an individual sample’s prediction, but individual confidence values may rise or fall due to the influence of signals from numerous other predictions (i.e., wisdom of crowds). Due to this fact, noisy signals misaligned with such ‘wisdom of crowds’, generally found in the correct signals, fail to raise the individual confidence values of wrong samples, despite attempts to increase them. Based on such findings, we filter out the samples whose confidence values are lower in the adapted model than in the original model, as they are likely to be noisy. Our method is widely applicable to existing TTA methods and improves their long-term adaptation performance in both image classification (e.g., 49.4% reduced error rates with TENT) and semantic segmentation (e.g., 11.7% gain in mIoU with TENT).

Identifying unexpected objects on roads in semantic segmentation (e.g., identifying dogs on roads) is crucial in safetycritical applications. Existing approaches use images of unexpected objects from external datasets or require additional training (e.g., retraining segmentation networks or training an extra network), which necessitate a non-trivial amount of labor intensity or lengthy inference time. One possible alternative is to use prediction scores of a pretrained network such as the max logits (i.e., maximum values among classes before the final softmax layer) for detecting such objects. However, the distribution of max logits of each predicted class is significantly different from each other, which degrades the performance of identifying unexpected objects in urban-scene segmentation. To address this issue, we propose a simple yet effective approach that standardizes the max logits in order to align the different distributions and reflect the relative meanings of max logits within each predicted class. Moreover, we consider the local regions from two different perspectives based on the intuition that neighboring pixels share similar semantic information. In contrast to previous approaches, our method does not utilize any external datasets or require additional training, which makes our method widely applicable to existing pretrained segmentation models. Such a straightforward approach achieves a new state-of-the-art performance on the publicly available Fishyscapes Lost & Found leader-board with a large margin. Our code is publicly available at this link1.

Image classification models tend to make decisions based on peripheral attributes of data items that have strong correlation with a target variable (i.e., dataset bias). These biased models suffer from the poor generalization capability when evaluated on unbiased datasets. Existing approaches for debiasing often identify and emphasize those samples with no such correlation (i.e., bias-conflicting) without defining the bias type in advance. However, such bias-conflicting samples are significantly scarce in biased datasets, limiting the debiasing capability of these approaches. This paper first presents an empirical analysis revealing that training with "diverse" bias-conflicting samples beyond a given training set is crucial for debiasing as well as the generalization capability. Based on this observation, we propose a novel feature-level data augmentation technique in order to synthesize diverse bias-conflicting samples. To this end, our method learns the disentangled representation of (1) the intrinsic attributes (i.e., those inherently defining a certain class) and (2) bias attributes (i.e., peripheral attributes causing the bias), from a large number of bias-aligned samples, the bias attributes of which have strong correlation with the target variable. Using the disentangled representation, we synthesize bias-conflicting samples that contain the diverse intrinsic attributes of bias-aligned samples by swapping their latent features. By utilizing these diversified bias-conflicting features during the training, our approach achieves superior classification accuracy and debiasing results against the existing baselines on both synthetic and real-world datasets.

Capturing human annotators' subjective responses in image annotation has become crucial as vision-based classifiers expand the range of application areas. While there has been significant progress in image annotation interface design in general, relatively little research has been conducted to understand how to elicit reliable and cost-efficient human annotation when the nature of the task includes a certain level of subjectivity. To bridge this gap, we aim to understand how different sampling methods in image batch labeling, a design that allows human annotators to label a batch of images simultaneously, can impact human annotation performances. In particular, we developed three different strategies in forming image batches: (1) uncertainty-based labeling (UL) that prioritizes images that a classifier predicts with the highest uncertainty, (2) certainty-based labeling (CL), a reverse strategy of UL, and (3) random, a baseline approach that randomly selects images. Although UL and CL solely select images to be labeled from a classifier's point of view, we hypothesized that human-side perception and labeling performance may also vary depending on the different sampling strategies. In our study, we observed that participants were able to recognize a different level of perceived cognitive load across three conditions (CL the easiest while UL the most difficult). We also observed a trade-off between annotation task effectiveness (CL and UL more reliable than random) and task efficiency (UL the most efficient while CL the least efficient). Based on the results, we discuss the implications of design and possible future research directions of image batch labeling.

Recently, there has been an increasing number of attempts to analyze sport activities through the combination of sports science and ICT technology.In the case of soccer, several leading companies have already developed tracking techniques for players to automatically acquire sports analysis data.However, the automatic extraction of event data for analyzing the sports games is limited to the level of academic research, and the field still depends on the manual work of professional analysts.This paper proposes a soccer event recognition technology based on pattern matching.As can be seen from the experimental results, it is possible to recognize various events much more accurately than the event recognition technology at academic research level.

We introduce a new type of organic user interface that displays a 3D robotic creature, ‘MoleBot’ to provide a ludic experience inspired by traditional board games. To ensure fluid motions of the molehills cast by the ‘MoleBot’, the table surface combines horizontal rigidity with the vertical flexibility of over 15,000 movable pins. Users are enabled to kinetically interact with this creature via a joystick or gestural commands. We conducted user study sessions with 12 participants and classified the observed spontaneous play activities into 4 distinct categories: (1) enjoying simple ludic experience, (2) competing in skills, (3) mimicking realworld sports and (4) playing with a companion. In addition, a focusgroup interview with six video scenarios was conducted to explore the idea of potential applications and it was suggested that the ‘MoleBot’ can be used in interactive board-gaming environments and kinetically informative tabletops.