I-INR: Iterative Implicit Neural Representations

An edge detection and surface reconstruction algorithm in which the smoothness is controlled spatially over the image space is presented. The values of parameters in the model are adaptively determined by an iterative refinement process; hence, the image-dependent parameters such as the optimum value of the regularization parameter or the filter size are eliminated. The algorithm starts with an oversmoothed regularized solution and iteratively refines the surface around discontinuities by using the structure exhibited in the error signal. The spatial control of smoothness is shown to resolve the conflict between detection and localization criteria of edge detection by smoothing the noise in continuous regions while preserving discontinuities. The performance of the algorithm is quantitatively and qualitatively evaluated on real and synthetic images, and it is compared with those of Marr-Hildreth and Canny edge detectors. >

As large language models (LLMs) are increasingly used to support learning, there is a growing need for a principled framework to guide the design of LLM-based tools and resources that are pedagogically effective and contextually responsive. This study proposes a framework by examining how prompt engineering can enhance the quality of chatbot responses to support middle school students’ scientific reasoning and argumentation. Drawing on learning theories and established frameworks for scientific argumentation, we employed a design-based research approach to iteratively refine system prompts and evaluate LLM-generated responses across diverse student input scenarios. Our analysis highlights how different prompt configurations affect the relevance and explanatory depth of chatbot feedback. We report findings from the iterative refinement process, along with an analysis of the quality of responses generated by each version of the chatbot. The outcomes indicate how different prompt configurations influence the coherence, relevance, and explanatory processes of LLM responses. The study contributes a set of critical design principles for developing theory-aligned prompts that enable LLM-based chatbots to meaningfully support students in constructing and revising scientific arguments. These principles offer broader implications for designing LLM applications across varied educational domains.

We consider the direct solution of sparse symmetric indefinite matrices. We develop new pivoting strategies that combine numerical pivoting and perturbation techniques. Then an iterative refinement process uses our approximate factorization to compute a solution. We show that our pivoting strategies are numerically robust, that few steps of iterative refinement are required, and that the factorization is significantly faster than with previous methods. Furthermore, we propose original approaches that are designed for parallel distributed factorization. A key point of our parallel implementation is the cheap and reliable estimation of the growth factor. This estimation is based on an approximation of the off-diagonal entries and does not require any supplementary messages.

One major challenge of QRS wave detection is the deleterious impact of inevitable noise in electrocardiogram (ECG) signals, which can deteriorate the detection performance of algorithms. Previous research has shown a nonlinear monostable potential‐based algorithm for enhancing the QRS complex in a noisy environment. Its main mechanism is the stochastic resonance effect, which transfers noise energy to information‐carrying ECG signals in the nonlinear potential. However, the QRS detection algorithm based on monostable stochastic resonance (MSR) exhibits a limited noise margin to maintain extremely good performance. To further improve noise robustness, we propose a multi‐parameter constrained bistable SR (MCBSR) module for the QRS detection algorithm without requiring a large amount of training data. The MCBSR module can maximize the ECG signal while suppressing noise when there is a QRS complex and minimizes the ECG signal otherwise. Using four publicly available databases for testing, the MCBSR‐based QRS detection algorithm has been proven to achieve superior performance among state‐of‐the‐art algorithms. Compared to the conventional nonlinear method, the MCBSR enables the QRS detection algorithm to have significantly stronger noise robustness under various noise types. Therefore, this study successfully provides a more excellent nonlinear potential and greatly enhances the noise robustness of the QRS detection algorithm in large noise environments. © 2025 The Author(s). IEEJ Transactions on Electrical and Electronic Engineering published by Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.

Image restoration driven by dual-scale prior.

Among the most popular tasks in computer vision, pedestrian re-identification (person re-identification, re-ID) is a technology that uses computer vision knowledge to determine if a particular person is present in an image or video sequence. Person re-ID has attracted substantial interest in the field of computer vision due to the growing demand for intelligent security. Person re-ID is often perturbed by factors such as occluded targets, unconstrained poses, and varying feature scales, and the application of convolutional neural network to extract global features of pedestrians for recognition often fails to produce suitable results. To obtain richer target features, it is advantageous to extract local features; Unfortunately, these features sometimes fail to align, which negatively impacts on the model. In this context, we propose a multi-branch neural network that combines global and local feature learning in order to obtain a more accurate representation of pedestrian features and enhance the recognition capacity of the model. We design a feature alignment module to further improve the performance of the model by implementing a spatial transformation of the feature maps generated by the network at different depths to achieve feature alignment. According to the experiments, the rank-1 and mAP indices on the Market-1501 dataset are 95.32% and 87.64%, respectively. the method achieves results on DukeMTMC-reID dataset with 88.21% rank-1 and 77.62% mAP. The results demonstrate that our method performs more efficiently against the state-of-the-art methods on two widely used person re-ID datasets.

Modern deep learning models achieve state-of-the-art results for many tasks in computer vision, such as image classification and segmentation. However, its adoption into high-risk applications, e.g. automated medical diagnosis systems, happens at a slow pace. One of the main reasons for this is that regular neural networks do not capture uncertainty. To assess uncertainty in classification, several techniques have been proposed casting neural network approaches in a Bayesian setting. Amongst these techniques, Monte Carlo dropout is by far the most popular. This particular technique estimates the moments of the output distribution through sampling with different dropout masks. The output uncertainty of a neural network is then approximated as the sample variance. In this paper, we highlight the limitations of such a variance-based uncertainty metric and propose an novel approach. Our approach is based on the overlap between output distributions of different classes. We show that our technique leads to a better approximation of the inter-class output confusion. We illustrate the advantages of our method using benchmark datasets. In addition, we apply our metric to skin lesion classification—a real-world use case—and show that this yields promising results.

Boosting feature selection for Neural Network based regression

In a previous paper, we introduced a deep learning neural network that should be able to detect the existence of very shallow periodic planetary transits in the presence of red noise. The network in that feasibility study would not provide any further details about the detected transits. The current paper completes this missing part. We present a neural network that tags samples that were obtained during transits. This is essentially similar to the task of identifying the semantic context of each pixel in an image—an important task in computer vision, called “semantic segmentation,” which is often performed by deep neural networks. The neural network we present makes use of novel deep learning concepts such as U-Nets, Generative Adversarial Networks, and adversarial loss. The resulting segmentation should allow further studies of the light curves that are tagged as containing transits. This approach toward the detection and study of very shallow transits is bound to play a significant role in future space-based transit surveys such as PLATO, which are specifically aimed to detect those extremely difficult cases of long-period shallow transits. Our segmentation network also adds to the growing toolbox of deep learning approaches that are being increasingly used in the study of exoplanets; but, so far mainly for vetting transits, rather than their initial detection.

Video Analytics analyzes the video content and adds brains to the eyes which means analytics to the camera. It extracts contents from the video by monitoring the video in real-time. Normal and Abnormal human activity detection using deep learning models is a challenging task in computer vision. The detection of the same will help in detecting crime scenes which will help in preventing treacherous actions Proposed method focuses on classifying normal activities for humans in real-time scenarios. The pre-processing technique for redundant frame detection, elimination, and training the model e?ciently using Convolutional Neural Network for classifying the activities is the main research contribution. The proposed method shows improvement in accuracy as compared to the reference method which can be further implemented for on edge embedded platforms for real-time applications

Object detection is a fundamental task in computer vision that involves identifying and localizing objects within images or video frames. This research focuses on implementing and evaluating the YOLOv5 (You Only Look Once version 5) model for real-time object detection. YOLOv5 is known for its efficiency, accuracy, and speed, making it a preferred choice for various applications such as autonomous driving, surveillance, and medical imaging. In this study, we explore the architecture, training process, and performance evaluation of YOLOv5 on benchmark datasets. The results demonstrate that YOLOv5 achieves high precision and recall, outperforming traditional object detection methods in terms of speed and accuracy. Keywords: Object Detection, YOLOv5, Deep Learning, Computer Vision, Real-Time Processing, Neural Networks.

Recent research has shown the initial success of sparse coding (Sc) in solving many computer vision tasks. Motivated by the fact that kernel trick can capture the nonlinear similarity of features, which helps in finding a sparse representation of nonlinear features, we propose kernel sparse representation (KSR). Essentially, KSR is a sparse coding technique in a high dimensional feature space mapped by an implicit mapping function. We apply KSR to feature coding in image classification, face recognition, and kernel matrix approximation. More specifically, by incorporating KSR into spatial pyramid matching (SPM), we develop KSRSPM, which achieves a good performance for image classification. Moreover, KSR-based feature coding can be shown as a generalization of efficient match kernel and an extension of Sc-based SPM. We further show that our proposed KSR using a histogram intersection kernel (HIK) can be considered a soft assignment extension of HIK-based feature quantization in the feature coding process. Besides feature coding, comparing with sparse coding, KSR can learn more discriminative sparse codes and achieve higher accuracy for face recognition. Moreover, KSR can also be applied to kernel matrix approximation in large scale learning tasks, and it demonstrates its robustness to kernel matrix approximation, especially when a small fraction of the data is used. Extensive experimental results demonstrate promising results of KSR in image classification, face recognition, and kernel matrix approximation. All these applications prove the effectiveness of KSR in computer vision and machine learning tasks.

As is well known, multibaseline phase unwrapping (PU) is put forward to overcome single-baseline PU in discontinuous-terrain-height estimation. This letter presents a refined algorithm based on the cluster analysis (CA)-based noise-robust efficient multibaseline PU algorithm proposed by H. Yu. The basic idea is to combine multiple interferometric synthetic aperture radar interferograms with different baseline lengths by a linear combination. The new interferograms after the linear combination increase the ambiguity heights. The number of resulting groups on the envelope of the intercept histogram is decreased and the distance between different intercept groups is widened. Compared with the conventional CA method, the significant advantage of the refined CA (RCA) algorithm is that it improves noise robustness when the intercept groups are densely distributed. The proposed RCA algorithm is validated using the simulated interferometric data. The results demonstrate that the noise robustness performance is better than that of the CA method.

High-speed automotive valve train design requires realistic models of the valve train. However, this frequently results in highly nonlinear systems with discontinuities and constraints. Optimality criteria and trade-offs for the designs are frequently performed through a process of simulation and iterative refinement. This paper presents CamOE, a cam design optimization package based on direct multiple shooting optimal control theory, incorporating structured sequential quadratic programming. The code allows the designer to incorporate the constraints of importance and to consider and synthesize appropriate optimality criteria. This allows him or her to synthesize the cam profile at the design stage without resorting to a tedious trial-and-error design process. This paper presents CamOE as a software environment that permits rapid feedback to the designer through the process of numerical experiments in specifying criteria and constraints on the automotive valve train.

Review and classification of AI-enabled COVID-19 CT imaging models based on computer vision tasks