arXiv subjects

Michał Romaszewski

Publications and source records attributed to Michał Romaszewski.

16 recordsLinked to original sources

arXiv2026

Toward Seasonal Guidelines for Robust Deep-Learning Sentinel-2 Building Detection in Different Area Types

Sentinel-2 imagery offers open access, global coverage, and frequent revisit times, making it attractive for practical building mapping at scale; however, its native 10m resolution makes building vs non-building classification challenging, particularly for small or sub-pixel buildings, and performance can vary with both seasonality and the heterogeneity of built-up environments. This paper introduces a Sentinel-2 building-detection framework designed to systematically quantify these effects and to support more formalised, practice-oriented model selection. We construct a dedicated multi-temporal Sentinel-2 dataset over the Warsaw region and derive binary ground-truth masks by rasterising official Polish topographic database (BDOT10k) building footprints onto the Sentinel-2 pixel grid. Using two established convolutional segmentation backbones (U-Net and DeepLabV3+), we first perform scene-specific fine-tuning to select a robust architecture and identify the best monthly models for L1C and L2A products separately. We then conduct cross-temporal inference by applying each best monthly model to all scenes, enabling an assessment of (i) which months provide favourable training and inference conditions, (ii) how performance transfers between seasons, (iii) the impact of processing level, and (iv) how these effects differ across built-up typologies. Based on these results, we provide practical guidance for routine Sentinel-2 building classification under varying acquisition periods and settlement characteristics.

cs.CV↗

arXiv2026

Enabling Cloud-Level Accuracy in Edge AI through IoT Data Preprocessing

Large language models (LLMs) offer a natural-language interface for interpreting Internet of Things (IoT) sensor data in smart environments; however, cloud deployment introduces latency, privacy, and connectivity concerns. Local LLMs can reduce these limitations, but compact edge-deployable models often show weaker numerical reasoning when raw sensor readings are provided directly. This paper investigates whether prompt-side preprocessing can improve the accuracy-latency trade-off of local LLMs for environmental monitoring. We propose a structured prompt construction framework that transforms raw air-quality and thermal-comfort measurements into progressively enriched textual representations: raw sensor values, threshold-aware descriptions, and compact environmental summary flags. The approach is evaluated using indoor Raspberry Pi/BME680 datasets from Tampere University and outdoor air-quality datasets from Helsinki, Katowice, and Warsaw. We construct a binary LLM query dataset covering air quality, thermal comfort, and joint environmental conditions, and evaluate five local and five cloud LLMs across three prompt variants and two inference modes, with and without chain-of-thought prompting. Results show that prompt enrichment substantially improves local-model accuracy. In No-CoT mode, local accuracy increases from 50.9% to 81.7% indoors and from 63.7% to 89.3% outdoors from the raw to the most enriched prompt. Local No-CoT inference is the fastest configuration, with mean latency close to 0.22 s, while CoT substantially increases inference time. These findings suggest that lightweight prompt-side preprocessing can narrow the local--cloud performance gap and support low-latency IoT analytics in smart environments.

cs.PF↗

arXiv2026

From Articles to Canopies: Knowledge-Driven Pseudo-Labelling for Tree Species Classification using LLM Experts

Hyperspectral tree species classification is challenging due to limited and imbalanced class labels, spectral mixing (overlapping light signatures from multiple species), and ecological heterogeneity (variability among ecological systems). Addressing these challenges requires methods that integrate biological and structural characteristics of vegetation, such as canopy architecture and interspecific interactions, rather than relying solely on spectral signatures. This paper presents a biologically informed, semi-supervised deep learning method that integrates multi-sensor Earth observation data, specifically hyperspectral imaging (HSI) and airborne laser scanning (ALS), with expert, ecological knowledge. The approach relies on biologically inspired pseudo-labelling over a precomputed canopy graph, yielding accurate classification at low training cost. In addition, ecological priors on species cohabitation are automatically derived from reliable sources using large language models (LLMs) and encoded as a cohabitation matrix with likelihoods of species occurring together. These priors are incorporated into the pseudo-labelling strategy, effectively introducing expert knowledge into the model. Experiments on a real-world forest dataset demonstrate 5.6% improvement over the best reference method. Expert evaluation of cohabitation priors reveals high accuracy with differences no larger than 15%.

cs.CV↗

arXiv2025

Semantic Depth Matters: Explaining Errors of Deep Vision Networks through Perceived Class Similarities

Understanding deep neural network (DNN) behavior requires more than evaluating classification accuracy alone; analyzing errors and their predictability is equally crucial. Current evaluation methodologies lack transparency, particularly in explaining the underlying causes of network misclassifications. To address this, we introduce a novel framework that investigates the relationship between the semantic hierarchy depth perceived by a network and its real-data misclassification patterns. Central to our framework is the Similarity Depth (SD) metric, which quantifies the semantic hierarchy depth perceived by a network along with a method of evaluation of how closely the network's errors align with its internally perceived similarity structure. We also propose a graph-based visualization of model semantic relationships and misperceptions. A key advantage of our approach is that leveraging class templates -- representations derived from classifier layer weights -- is applicable to already trained networks without requiring additional data or experiments. Our approach reveals that deep vision networks encode specific semantic hierarchies and that high semantic depth improves the compliance between perceived class similarities and actual errors.

cs.CV↗

arXiv2025

Quantum-aware Transformer model for state classification

Entanglement is a fundamental feature of quantum mechanics, playing a crucial role in quantum information processing. However, classifying entangled states, particularly in the mixed-state regime, remains a challenging problem, especially as system dimensions increase. In this work, we focus on bipartite quantum states and present a data-driven approach to entanglement classification using transformer-based neural networks. Our dataset consists of a diverse set of bipartite states, including pure separable states, Werner entangled states, general entangled states, and maximally entangled states. We pretrain the transformer in an unsupervised fashion by masking elements of vectorized Hermitian matrix representations of quantum states, allowing the model to learn structural properties of quantum density matrices. This approach enables the model to generalize entanglement characteristics across different classes of states. Once trained, our method achieves near-perfect classification accuracy, effectively distinguishing between separable and entangled states. Compared to previous Machine Learning, our method successfully adapts transformers for quantum state analysis, demonstrating their ability to systematically identify entanglement in bipartite systems. These results highlight the potential of modern machine learning techniques in automating entanglement detection and classification, bridging the gap between quantum information theory and artificial intelligence.

quant-ph↗

arXiv2024

Efficient Numerical Calibration of Water Delivery Network Using Short-Burst Hydrant Trials

Calibration is a critical process for reducing uncertainty in Water Distribution Network Hydraulic Models (WDN HM). However, features of certain WDNs, such as oversized pipelines, lead to shallow pressure gradients under normal daily conditions, posing a challenge for effective calibration. This study proposes a calibration methodology using short hydrant trials conducted at night, which increase the pressure gradient in the WDN. The data is resampled to align with hourly consumption patterns. In a unique real-world case study of a WDN zone, we demonstrate the statistically significant superiority of our method compared to calibration based on daily usage. The experimental methodology, inspired by a machine learning cross-validation framework, utilises two state-of-the-art calibration algorithms, achieving a reduction in absolute error of up to 45% in the best scenario.

eess.SP↗

arXiv2024

`Just One More Sensor is Enough' -- Iterative Water Leak Localization with Physical Simulation and a Small Number of Pressure Sensors

In this article, we propose an approach to leak localisation in a complex water delivery grid with the use of data from physical simulation (e.g. EPANET software). This task is usually achieved by a network of multiple water pressure sensors and analysis of the so-called sensitivity matrix of pressure differences between the network's simulated data and actual data of the network affected by the leak. However, most algorithms using this approach require a significant number of pressure sensors -- a condition that is not easy to fulfil in the case of many less equipped networks. Therefore, we answer the question of whether leak localisation is possible by utilising very few sensors but having the ability to relocate one of them. Our algorithm is based on physical simulations (EPANET software) and an iterative scheme for mobile sensor relocation. The experiments show that the proposed system can equalise the low number of sensors with adjustments made for their positioning, giving a very good approximation of leak's position both in simulated cases and real-life example taken from BattLeDIM competition L-Town data.

eess.SY↗

arXiv2024

Through the Thicket: A Study of Number-Oriented LLMs derived from Random Forest Models

Large Language Models (LLMs) have shown exceptional performance in text processing. Notably, LLMs can synthesize information from large datasets and explain their decisions similarly to human reasoning through a chain of thought (CoT). An emerging application of LLMs is the handling and interpreting of numerical data, where fine-tuning enhances their performance over basic inference methods. This paper proposes a novel approach to training LLMs using knowledge transfer from a random forest (RF) ensemble, leveraging its efficiency and accuracy. By converting RF decision paths into natural language statements, we generate outputs for LLM fine-tuning, enhancing the model's ability to classify and explain its decisions. Our method includes verifying these rules through established classification metrics, ensuring their correctness. We also examine the impact of preprocessing techniques on the representation of numerical data and their influence on classification accuracy and rule correctness

cs.CL↗

arXiv2024

Data structure > labels? Unsupervised heuristics for SVM hyperparameter estimation

Classification is one of the main areas of pattern recognition research, and within it, Support Vector Machine (SVM) is one of the most popular methods outside of field of deep learning -- and a de-facto reference for many Machine Learning approaches. Its performance is determined by parameter selection, which is usually achieved by a time-consuming grid search cross-validation procedure (GSCV). That method, however relies on the availability and quality of labelled examples and thus, when those are limited can be hindered. To address that problem, there exist several unsupervised heuristics that take advantage of the characteristics of the dataset for selecting parameters instead of using class label information. While an order of magnitude faster, they are scarcely used under the assumption that their results are significantly worse than those of grid search. To challenge that assumption, we have proposed improved heuristics for SVM parameter selection and tested it against GSCV and state of the art heuristics on over 30 standard classification datasets. The results show not only its advantage over state-of-art heuristics but also that it is statistically no worse than GSCV.

cs.LG↗

arXiv2022

Improving Autoencoder Training Performance for Hyperspectral Unmixing with Network Reinitialisation

Neural networks, in particular autoencoders, are one of the most promising solutions for unmixing hyperspectral data, i.e. reconstructing the spectra of observed substances (endmembers) and their relative mixing fractions (abundances), which is needed for effective hyperspectral analysis and classification. However, as we show in this paper, the training of autoencoders for unmixing is highly dependent on weights initialisation; some sets of weights lead to degenerate or low-performance solutions, introducing negative bias in the expected performance. In this work, we experimentally investigate autoencoders stability as well as network reinitialisation methods based on coefficients of neurons' dead activations. We demonstrate that the proposed techniques have a positive effect on autoencoder training in terms of reconstruction, abundances and endmembers errors.

eess.IV↗

arXiv2021

A Dataset for Evaluating Blood Detection in Hyperspectral Images

The sensitivity of imaging spectroscopy to haemoglobin derivatives makes it a promising tool for detecting blood. However, due to complexity and high dimensionality of hyperspectral images, the development of hyperspectral blood detection algorithms is challenging. To facilitate their development, we present a new hyperspectral blood detection dataset. This dataset, published in accordance to open access mandate, consist of multiple detection scenarios with varying levels of complexity. It allows to test the performance of Machine Learning methods in relation to different acquisition environments, types of background, age of blood and presence of other blood-like substances. We explored the dataset with blood detection experiments. We used hyperspectral target detection algorithm based on the well-known Matched Filter detector. Our results and their discussion highlight the challenges of blood detection in hyperspectral data and form a reference for further works.

cs.CV↗

arXiv2020

Hyperspectral classification of blood-like substances using machine learning methods combined with genetic algorithms in transductive and inductive scenarios

This study is focused on applying genetic algorithms (GA) to model and band selection in hyperspectral image classification. We use a forensic-inspired data set of seven hyperspectral images with blood and five visually similar substances to test GA-optimised classifiers in two scenarios: when the training and test data come from the same image and when they come from different images, which is a more challenging task due to significant spectra differences. In our experiments we compare GA with a classic model optimisation through grid search. Our results show that GA-based model optimisation can reduce the number of bands and create an accurate classifier that outperforms the GS-based reference models, provided that during model optimisation it has access to examples similar to test data. We illustrate this with experiment highlighting the importance of a validation set.

cs.CV↗

arXiv2020

Comparison of surface thermal patterns of horses and donkeys in IRT images

Infrared thermography (IRT) is a valuable diagnostic tool in equine veterinary medicine however, little is known about its application in donkeys. The aim was to find patterns in thermal images of donkeys and horses, and determine if these patterns share similarities. The study was carried out on 18 donkeys and 16 horses. All equids underwent thermal imaging with an infrared camera and measuring the skin thickness and hair coat length. On the class maps of each thermal image, 15 regions of interest (ROIs) were annotated and then combined into 10 groups of ROIs (GORs). The existence of statistically significant differences between surface temperatures in GORs was tested both `globally' for all animals of a given species and `locally' for each animal. Two special cases of animals that differ from the rest were also discussed. Our results indicated that the majority of thermal patterns are similar for both species however, average surface temperatures in horses are higher than in donkeys. It may be related to differences in the skin and hair coat. We concluded, the patterns of both species are associated with GORs, rather than an individual ROI, with higher uniformity of donkeys patterns.

q-bio.QM↗

arXiv2018

Band selection with Higher Order Multivariate Cumulants for small target detection in hyperspectral images

In the small target detection problem a pattern to be located is on the order of magnitude less numerous than other patterns present in the dataset. This applies both to the case of supervised detection, where the known template is expected to match in just a few areas and unsupervised anomaly detection, as anomalies are rare by definition. This problem is frequently related to the imaging applications, i.e. detection within the scene acquired by a camera. To maximize available data about the scene, hyperspectral cameras are used; at each pixel, they record spectral data in hundreds of narrow bands. The typical feature of hyperspectral imaging is that characteristic properties of target materials are visible in the small number of bands, where light of certain wavelength interacts with characteristic molecules. A target-independent band selection method based on statistical principles is a versatile tool for solving this problem in different practical applications. Combination of a regular background and a rare standing out anomaly will produce a distortion in the joint distribution of hyperspectral pixels. Higher Order Cumulants Tensors are a natural `window' into this distribution, allowing to measure properties and suggest candidate bands for removal. While there have been attempts at producing band selection algorithms based on the 3 rd cumulant's tensor i.e. the joint skewness, the literature lacks a systematic analysis of how the order of the cumulant tensor used affects effectiveness of band selection in detection applications. In this paper we present an analysis of a general algorithm for band selection based on higher order cumulants. We discuss its usability related to the observed breaking points in performance, depending both on method order and the desired number of bands. Finally we perform experiments and evaluate these methods in a hyperspectral detection scenario.

cs.CV↗

arXiv2013

Compression of animated 3D models using HO-SVD

This work presents an analysis of Higher Order Singular Value Decomposition (HO-SVD) applied to lossy compression of 3D mesh animations. We describe strategies for choosing a number of preserved spatial and temporal components after tensor decomposition. Compression error is measured using three metrics (MSE, Hausdorff, MSDM). Results are compared with a method based on Principal Component Analysis (PCA) and presented on a set of animations with typical mesh deformations.

cs.GR↗

arXiv2013

Natural hand gestures for human identification in a Human-Computer Interface

The goal of this work is the identification of humans based on motion data in the form of natural hand gestures. In this paper, the identification problem is formulated as classification with classes corresponding to persons' identities, based on recorded signals of performed gestures. The identification performance is examined with a database of twenty-two natural hand gestures recorded with two types of hardware and three state-of-art classifiers: Linear Discrimination Analysis (LDA), Support Vector machines (SVM) and k-Nearest Neighbour (k-NN). Results show that natural hand gestures allow for an effective human classification.

cs.HC↗