Dataset distillation (DD) aims to generate a compact synthetic dataset that enables efficient training of neural networks while maintaining performance comparable to that achieved with the original dataset. However, existing methods often suffer from two main limitations. They either rely on computationally intensive iterative optimization procedures or depend heavily on architecture-specific designs. These issues limit their practicality for large-scale datasets and hinder generalization across different model architectures. To overcome these challenges, recent research has explored the use of diffusion models as an architecture-agnostic approach to dataset distillation, offering improved scalability and generalization for large-scale datasets across diverse model architectures. While diffusion-based dataset distillation methods have shown considerable potential, several challenges remain. Notably, certain approaches exhibit a distributional mismatch between the pre-trained diffusion model and the target dataset, which can adversely affect the fidelity and representativeness of the generated samples. Others require substantial fine-tuning to achieve high fidelity, which negates the benefits of architectural flexibility. In this work, we propose a new diffusion-based dataset distillation framework that effectively preserves the characteristics of the original dataset without requiring any fine-tuning. Our method employs adaptive sampling and repulsion regularization to enhance both the fidelity and diversity of generated samples. As a result, the proposed approach outperforms state-of-the-art distillation methods across a wide range of datasets and model architectures.

Pruning is a highly effective method for reducing the size of neural networks with negligible impact on their average performance. However, recent studies have revealed that pruning actually amplifies the bias in the models, leading to decreased performance for underrepresented groups. To address this issue, we first analyze the impact of pruning on the confidence of each sample and introduce Accumulated Confidence (AC). AC is a proxy that facilitates the identification of bias-conflicting and bias-aligned samples without relying on group annotations. We then propose a debiasing algorithm, which is called DEbiasing Network through Pruning (DENP). DENP utilizes AC to mitigate bias within the network. Even without bias information, DENP exhibits remarkable debiasing performance on varying levels of sparsity, effectively mitigating the bias-exacerbating property of pruning and resulting in both sparse and debiased neural networks. Moreover, even when compared with state-of-the-art debiasing baselines under identical conditions, the DENP still achieves the best performance on multiple benchmark datasets, demonstrating its superior debiasing capabilities.

As the use of machine learning models has increased, numerous studies have aimed to enhance fairness. However, research on the intersection of fairness and explainability remains insufficient, leading to potential issues in gaining the trust of actual users. Here, we propose a novel module that constructs a fair latent space, enabling faithful explanation while ensuring fairness. The fair latent space is constructed by disentangling and redistributing labels and sensitive attributes, allowing the generation of counterfactual explanations for each type of information. Our module is attached to a pretrained generative model, transforming its biased latent space into a fair latent space. Additionally, since only the module needs to be trained, there are advantages in terms of time and cost savings, without the need to train the entire generative model. We validate the fair latent space with various fairness metrics and demonstrate that our approach can effectively provide explanations for biased decisions and assurances of fairness.

Despite their impressive generative capabilities, text-to-image diffusion models often memorize and replicate training data, prompting serious concerns over privacy and copyright. Recent work has attributed this memorization to an attraction basin—a region where applying classifier-free guidance (CFG) steers the denoising trajectory toward memorized outputs—and has proposed deferring CFG application until the denoising trajectory escapes this basin. However, such delays often result in non-memorized images that are poorly aligned with the input prompts, highlighting the need to promote earlier escape so that CFG can be applied sooner in the denoising process. In this work, we show that the initial noise sample plays a crucial role in determining when this escape occurs. We empirically observe that different initial samples lead to varying escape times. Building on this insight, we propose two mitigation strategies that adjust the initial noise—either collectively or individually—to find and utilize initial samples that encourage earlier basin escape. These approaches significantly reduce memorization while preserving image-text alignment.

Models trained with empirical risk minimization (ERM) are prone to be biased towards spurious correlations between target labels and bias attributes, which leads to poor performance on data groups lacking spurious correlations. It is particularly challenging to address this problem when access to bias labels is not permitted. To mitigate the effect of spurious correlations without bias labels, we first introduce a novel training objective designed to robustly enhance model performance across all data samples, irrespective of the presence of spurious correlations. From this objective, we then derive a debiasing method, Disagreement Probability based Resampling for debiasing (DPR), which does not require bias labels. DPR leverages the disagreement between the target label and the prediction of a biased model to identify bias-conflicting samples—those without spurious correlations—and upsamples them according to the disagreement probability. Empirical evaluations on multiple benchmarks demonstrate that DPR achieves state-of-the-art performance over existing baselines that do not use bias labels. Furthermore, we provide a theoretical analysis that details how DPR reduces dependency on spurious correlations.

This paper is the first to attempt differentially private (DP) topological data analysis (TDA), producing near-optimal private persistence diagrams. We analyze the sensitivity of persistence diagrams in terms of the bottleneck distance, and we show that the commonly used Cech complex has sensitivity that does not decrease as the sample size $n$ increases. This makes it challenging for the persistence diagrams of Cech complexes to be privatized. As an alternative, we show that the persistence diagram obtained by the $L^1$-distance to measure (DTM) has sensitivity $O(1/n)$. Based on the sensitivity analysis, we propose using the exponential mechanism whose utility function is defined in terms of the bottleneck distance of the $L^1$-DTM persistence diagrams. We also derive upper and lower bounds of the accuracy of our privacy mechanism; the obtained bounds indicate that the privacy error of our mechanism is near-optimal. We demonstrate the performance of our privatized persistence diagrams through simulations as well as on a real data set tracking human movement.

One of the main limitations in the field of audio signal processing is the lack of large public datasets with audio representations and high-quality annotations due to restrictions of copyrighted commercial music. We present Melon Playlist Dataset, a public dataset of mel-spectrograms for 649,091 tracks and 148,826 associated playlists annotated by 30,652 different tags. All the data is gathered from Melon, a popular Korean streaming service. The dataset is suitable for music information retrieval tasks, in particular, auto-tagging and automatic playlist continuation. Even though the latter can be addressed by collaborative filtering approaches, audio provides opportunities for research on track suggestions and building systems resistant to the cold-start problem, for which we provide a baseline. Moreover, the playlists and the annotations included in the Melon Playlist Dataset make it suitable for metric learning and representation learning.

This work presents energy harvesting techniques from low-voltage current used to prevent galvanic corrosion between a metallic structure and a permanent copper/copper sulfate (Cu/CuSO4) reference electrode. Supercapacitors are adopted to compensate for or overcome the limitations of batteries. Then, a boost converter is used to convert the low voltage levels of galvanic corrosion to that needed by the complementary metal oxide semiconductor (CMOS) technologies used for the wireless sensor systems. Experimental results show that our proposed harvesting schemes significantly reduce the overhead of the charging circuitry, which enables nearly full charging of supercapacitors of up to 350F under the low power conditions of 3mW (i.e., 3mA at 1,V). More importantly, our system enables maintenance-free operation of remote-monitoring cathodic protection (RMCP) systems in harsh environments, where sunlight or wind power may be unavailable or unpredictable.

Power-transmission loss can be a severe problem for low-power embedded systems organized in a daisy-chain topology. The loss can be so high that it can result in failure to power the load in the first place. The first contribution of this article is a recursive algorithm for solving the transmission current on each segment of the daisy chain at a given supply voltage. It enables solving not only the transmission loss but also reports infeasible configurations if the voltage is too low. Using this core algorithm, our second contribution is to find energy-efficient configurations that use local energy buffers (LEBs) to eliminate peak load on the bus without relying on high voltage. Experimental results confirm that our proposed techniques significantly reduce the total energy consumption and enable the deployed system to operate for significantly longer.

The acquisition of surround-view panoramas using a single hand-held or head-worn camera relies on robust real-time camera orientation tracking and relocalization. This paper presents robust methodology and evaluation for camera orientation relocalization, using virtual keyframes for online environment map construction. In the case of tracking loss, incoming camera frames are matched against known-orientation keyframes to re-estimate camera orientation. Instead of solely using real keyframes from incoming video, the proposed approach employs virtual keyframes which are distributed strategically within completed portions of an environment map. To improve tracking speed, we introduce a new variant of our system which carries out relocalization only when tracking fails and uses inexpensive image-patch descriptors. We compare different system variants using three evaluation methods to show that the proposed system is useful in a practical sense. To improve relocalization robustness against lighting changes in indoor and outdoor environments, we propose a new approach based on illumination normalization and saturated area removal. We examine the performance of our solution over several indoor and outdoor video sequences, evaluating relocalization rates based on ground truth from a pan-tilt unit.

We present an analysis of four orientation tracking systems used for construction of environment maps. We discuss the analysis necessary to determine the robustness of tracking systems in general. Due to the difficulty inherent in collecting user evaluation data, we then propose a metric which can be used to obtain a relative estimate of these values. The proposed metric will still require a set of input videos with an associated distance to ground truth, but not an additional user evaluation.

The acquisition of surround-view panoramas using a single hand-held or head-worn camera relies on robust real-time camera orientation tracking. In absence of robust tracking recovery methods, the complete acquisition process has to be re-started when tracking fails. This paper presents methodology for camera orientation relocalization, using virtual keyframes for online environment map construction. Instead of relying on real keyframes from incoming video, the proposed approach enables camera orientation relocalization by employing virtual keyframes which are distributed strategically within an environment map. We discuss our insights about a suitable number and distribution of virtual keyframes, as suggested by our experiments on virtual keyframe generation and orientation relocalization. After a shading correction step, we relocalize camera orientation in real-time by comparing the current camera frame to virtual keyframes. While expanding the captured environment map, we continue to simultaneously generate virtual keyframes within the completed portion of the map, as descriptors to estimate camera orientation. We implemented our camera orientation relocalizer with the help of a GPU fragment shader for real-time application, and evaluated the speed and accuracy of the proposed approach.

This paper presents an online 3D modeling and tracking methodology that uses aerial photographs for mobile augmented reality. Instead of relying on models which are created in advance, the system generates a 3D model for a real building on the fly by combining frontal and aerial views with the help of an optical sensor, an inertial sensor, a GPS unit and a few mouse clicks. A user’s initial pose is estimated using an aerial photograph, which is retrieved from a database according to the user’s GPS coordinates, and an inertial sensor which measures pitch. To track the user’s position and orientation in real-time, feature-based tracking is carried out based on salient points on the edges and the sides of a building the user is keeping in view. We implemented camera pose estimators using both a least squares and an unscented Kalman filter (UKF) approach. The UKF approach results in more stable and reliable vision-based tracking. We evaluate the speed and accuracy of both approaches, and we demonstrate the usefulness of our computations as important building blocks for an Anywhere Augmentation scenario.

In this paper, we propose a new method for generating an image-based 3D panoramic virtual environment (VE). The panoramic VE is generated using 3D depth information estimated from rotating two multiview cameras. Even though conventional 2D image-based mosaicking methods provide a wide view, they have limitations in providing a user with a navigation-enabled virtual environment. In order to resolve such obstacles, we first estimate the depth of the scene using two calibrated multiview cameras and then stitch 3D point clouds instead of images. By rotating two cameras using a turn-table it enables users to navigate the resulting 3D virtual environment with HMD.

A fundamental issue in content-based image retrieval is how to select image features that can represent image contents appropriately. A multi-scale color clustering algorithm based on human perceptual properties of color images is proposed for image retrieval. The multi-scale clustering algorithm is an unsupervised clustering method that utilizes the perceptual uniformity property in the (p,q) color space. The proposed color clustering algorithm produces a small set of representative color vectors for each image that capture color properties of the image, and a set of correlogram values that contain the spatial information of the image.