Przeglądaj {{ collection }} wg Autor "Tomczyk, Arkadiusz"

Teraz wyświetlane 1 - 7 z 7

Active partition based medical image understanding with self-organised competitive spatch eduction
(Wydawnictwo Politechniki Łódzkiej, 2010) Pryczek, Michał; Tomczyk, Arkadiusz; Szczepaniak, Piotr
Medical Image Understanding is a recently defined semantic oriented image recognition task. Its specific requirements, highlighting complex characteristics of recognised objects as well as indispensable use of human-level expert knowledge almost every step of data processing sets new requirements for implemented algorithms. This paper focuses on linguistic image description method, designed to segment low level, semantically coherent image regions and mine adjacency relations among them. Example method results on medical images are presented to specify some methods properties.
Application of active contours with expert knowledge to heart ventricle segmentation
(Wydawnictwo Politechniki Łódzkiej, 2013) Tomczyk, Arkadiusz
Automatic heart ventricle segmentation in CT heart images can be an element of system supporting pulmonary embolism diagnosis. To solve that problem in this paper an application of two classical active contour models, snakes and geometric active contours, is proposed. The prepared implementation uses the unified model of those techniques which allows to define forces acting upon a contour only once. The nature of the images causes that the process of force construction requires additional expert knowledge since using only the information visible in the image satisfactory results cannot be obtained.
Detection of line segments
(Wydawnictwo Politechniki Łódzkiej, 2014) Tomczyk, Arkadiusz
In the paper a method of line segment detection in the images is presented. It bases on existing LSD approach which was designed to be a parameterless technique dedicated to analysis of real world scenes. In consequence it encounters problems with other types of images, e.g. medical images, where the characteristic of the structures may be completely different. The method proposed in this work allows to tune its parameters for that characteristic and thus allows to achieve satisfactory results also for medical data. The effect of the approach is illustrated with mammographic images.
Energy Dissipation Anomalies in Buildings
(Wydawnictwo Politechniki Łódzkiej, 2023) Morawski, Michał; Tomczyk, Arkadiusz; Idaczyk, Maciej
Improvement of Attention Mechanism Explainability in Prediction of Chemical Molecules’ Properties
(Wydawnictwo Politechniki Łódzkiej, 2023) Durys, Bartosz; Tomczyk, Arkadiusz
In this paper, the analysis of selected graph neural network operators is presented. The classic Graph Convolutional Network (GCN) was compared with methods containing trainable attention coefficients: Graph Attention Network (GAT) and Graph Transformer (GT). Moreover, which is an original contribution of this work, training of GT was modified with an additional loss function component enabling easier explainability of the produced model. The experiments were conducted using datasets with chemical molecules where both classification and regression tasks are considered. The results show that additional constraint not only does not make the results worse but, in some cases, it improves predictions.
Spatch based active partitions with linguistically formulated energy
(Wydawnictwo Politechniki Łódzkiej, 2010) Tomczyk, Arkadiusz; Pryczek, Michał; Walczak, Stanisław; Jojczyk, Konrad; Szczepaniak, Piotr
The present paper shows the method of cognitive hierarchical active partitions that can be applied to creation of automatic image understanding systems. The approach, which stems from active contours techniques, allows one to use not only the knowledge contained in an image, but also any additional expert knowledge. Special emphasis is put on the effcient way of knowledge retrieval, which could minimise the necessity to render information expressed in a natural language into a description convenient for recognition algorithms and machine learning.
Synergy of Convolutional Neural Networks and Geometric Active Contours
(Wydawnictwo Politechniki Łódzkiej, 2021) Tomczyk, Arkadiusz; Pankiv, Oleksandr; Szczepaniak, Piotr S.
Hybrid approach to machine learning techniques could potentially provide improvements in image segmentation results. In this paper, a model of cooperation of convolutional neural networks and geometric active contours is proposed and developed. The novelty of the approach lies in combining deep neural networks and active contour model in order to improve CNN output results. The method is examined on the image segmentation task and applied to the detection and extraction of nuclei of HL60 cell line. The model had been tested on both 2-D and 3-D images. Because of feature learning characteristics of convolutional neural networks, the proposed solution should perform well in multiple scenarios and can be considered generic.