Wyniki wyszukiwania - Biblioteka Nauki

1

Analysis of the possibility of using the singular value decomposition in image compression

100%

Łukasik E. , Łabuć E.

Applied Computer Science

|

2022

|

tom Vol. 18, no 4

53--67

EN

In today’s highly computerized world, data compression is a key issue to minimize the costs associated with data storage and transfer. In 2019, more than 70% of the data sent over the network were images. This paper analyses the feasibility of using the SVD algorithm in image compression and shows that it improves the efficiency of JPEG and JPEG2000 compression. Image matrices were decomposed using the SVD algorithm before compression. It has also been shown that as the image dimensions increase, the fraction of eigenvalues that must be used to reconstruct the image in good quality decreases. The study was carried out on a large and diverse set of images, more than 2500 images were examined. The results were analyzed based on criteria typical for the evaluation of numerical algorithms operating on matrices and image compression: compression ratio, size of compressed file, MSE, number of bad pixels, complexity, numerical stability, easiness of implementation.

2

Efficiency comparison of networks in handwritten latin characters recognition with diacritics

100%

Łukasik E. , Flis W.

Applied Computer Science

|

2023

|

tom Vol. 19, no 4

88--102

EN

The aim of the article is to analyze and compare the performance and accuracy of architectures with a different number of parameters on the example of a set of handwritten Latin characters from the Polish Handwritten Characters Database (PHCD). It is a database of handwriting scans containing letters of the Latin alphabet as well as diacritics characteristic of the Polish language. Each class in the PHCD dataset contains 6,000 scans for each character. The research was carried out on six proposed architectures and compared with the architecture from the literature. Each of the models was trained for 50 epochs, and then the accuracy of prediction was measured on a separate test set. The experiment thus constructed was repeated 20 times for each model. Accuracy, number of parameters and number of floating-point operations performed by the network were compared. The research was conducted on subsets such as uppercase letters, lowercase letters, lowercase letters with diacritics, and a subset of all available characters. The relationship between the number of parameters and the accuracy of the model was indicated. Among the examined architectures, those that significantly improved the prediction accuracy at the expense of a larger network size were selected, and a network with a similar prediction accuracy as the base one, but with twice as many model parameters was selected.

3

Model Faster R-CNN uczony na syntetycznych obrazach

100%

Łach B. , Łukasik E.

Journal of Computer Sciences Institute

|

2020

|

tom Vol. 17

401--404

PL

Uczenie maszynowe wymaga opisu danych przez człowieka. Opisywanie zbioru danych ręcznie jest bardzo czasochłonne. W artykule zbadano jak model uczył się na zdjęciach sztucznie wytworzonych, z jak najmniejszym udziałem człowieka przy opisywaniu danych. Sprawdzono jaki wpływ miało zastosowanie augmentacji i progresywnego rozmiaru zdjęcia przy treningu modelu na syntetycznym zbiorze. Model osiągnął nawet o 3,35% wyższą średnią precyzję na syntetycznym zbiorze danych przy zastosowaniu treningów z rosnącą rozdzielczością. Augmentacje poprawiły jakość detekcji na rzeczywistych zdjęciach. Wytwarzanie sztucznie danych treningowych ma duży wpływ na przyśpieszenie przygotowania treningów, ponieważ nie wymaga tak dużych nakładów ludzkich, jak klasyczne uczenie modeli z danymi opisanymi przez człowieka.

EN

Machine learning requires a human description of the data. The manual dataset description is very time consuming. In this article was examined how the model learns from artificially created images, with the least human participation in describing the data. It was checked how the model learned on artificially produced images with augmentations and progressive image size. The model has achieve up to 3.35 higher mean average precision on syntetic dataset in the training with increasing images resolution. Augmentations improved the quality of detection on real photos. The production of artificially generated training data has a great impact on the acceleration of prepare training, because it does not require as much human resources as normal learning process.

4

Examination of text's lexis using a Polish dictionary

100%

Voitovych R. , Łukasik E.

Journal of Computer Sciences Institute

|

2021

|

tom Vol. 21

316--323

EN

This paper presents an approach to compare and classify books written in the Polish language by comparing their lexis fields. Books can be classified by their features, such as literature type, literary genre, style, author, etc. Using a preas-sembled dictionary and Jaccard index, the authors managed to prove a lexical likeness for books. Further analysis with the PAM clustering algorithm presented a lexical connection between books of the same type or author. Analysis of values of similarities of any particular field on one side and some anomalous tendencies in other cases suggest that recognition of other features is possible. The method presented in this article allows to draw conclusions about the con-nection between any arbitrary books based solely on their vocabulary.

PL

Artykuł prezentuje metodę porównania i klasyfikacji książek napisanych w języku polskim na podstawie ich leksyki. Książki można dzielić, korzystając z ich cech, np. rodzaju literatury, gatunku literackiego, stylu, autora itp. Korzystając ze skompilowanego słownika i indeksu Jaccarda, udowodniona została hipoteza dotycząca podobieństwa książek rozpatrywanego pod kątem ich leksyki. Kolejna analiza za pomocą algorytmu klastrowego PAM wskazuje na związek leksykalny pomiędzy książkami jednego rodzaju literatury lub autora. Analiza wartości współczynników poszczególnych obszarów z jednej strony i anomalia w zachowaniu w niektórych przypadkach sugeruje, że wyodrębnienie kolejnych cech jest możliwe. Metoda przedstawiona w tym artykule pozwala wyciągać wnioski o relacjach między książkami, korzystając wyłącznie z ich słownictwa.

5

Analiza porównawcza wybranych programów do optycznego rozpoznawania tekstu

100%

Łukasik E. , Zientarski T.

Journal of Computer Sciences Institute

|

2018

|

tom Vol. 7

191--194

PL

Celem artykułu jest porównanie trzech programów do optycznego rozpoznawania teksu. Zdefiniowany został problem optycznego rozpoznawania tekstu i przedstawione główne jego zastosowania. Opisano działanie tej technologii i krótko scharakteryzowano najważniejsze dostępne na rynku programy realizujące omawiane zagadnienie. Następnie poddano testom wybrane programy wykorzystując dwie próbki pisma maszynowego w języku polskim. Określono szybkość procesu rozpoznawania tekstu. Poprawność rozpoznania znaków i wyrazów w analizowanym tekście została także określona.

EN

The aim of the article is to compare three programs for the optical text recognition. The problem of the optical text recognition has been defined. Next, briefly the functionality of this technology was described. The most important programs realizing the discussed problem were also characterized. The selected programs were tested using two samples of machine writing in Polish. The speed of the text recognition process was determined. The correctness of characters and words recognition in the analyzed text was also specified.

6

Szybkość uczenia czy dokładność predykcji? Analiza porównawcza szkieletów programistycznych do sztucznej inteligencji

80%

Zdeb K. , Żukiewicz P. , Łukasik E.

Journal of Computer Sciences Institute

|

2022

|

tom Vol. 24

172--175

PL

Celem artykułu jest analiza szkieletów aplikacji do sztucznej inteligencji. Zbadane zostały: skuteczność, czasochłonność oraz ilość potrzebnych zasobów. Dla każdego frameworka stworzono modele regresji liniowej, lasów losowych i k najbliższych sąsiadów. Dane uczące to zbiory danych zawierające informację o diamencie oraz jego cenie. Każdy model miał za zadanie nauczyć się cen diamentów, a następnie dokonać predykcji w zależności od ich konkretnych cech tj. szlif, kolor, objętość. Dane uczące zostały podzielone na zbiory o różnej wielkości dzięki czemu można było zaobserwować zmianę w modelu w zależności od liczby danych treningowych. Z trzech przebadanych szkieletów programistycznych do uczenia maszynowego TensorFlow wykazał się największą skutecznością, a SciKit-Learn najkrótszym czasem dokonywania predykcji

EN

The purpose of the article is to analyze frameworks for artificial intelligence applications. In particular, the effectiveness, time-consumption and resources requirement. Linear regression, random forests and k nearest neighbors models were created for each framework. The learning data is a dataset containing informations about diamonds and their prices. Each model was designed to learn diamonds’ prices and then make a prediction depending on its specific characteristics such as cut, color, and volume. The learning data was divided into sets of different sizes to show changes in a model depending on the amount of training data. Out of the three machine learning frameworks tested, TensorFlow proved to be the most accurate and SciKit-Learn the fastest

7

Relationship of lumbar-hip kinematics during trunk flexion and sex, body mass index, and self-reported energy expenditure: a cross-sectional analysis

61%

Zawadka M. , Smołka J. , Skublewska-Paszkowska M. , Łukasik E. , Zieliński G. , Gawda P.

Acta of Bioengineering and Biomechanics

|

2023

|

tom Vol. 25, no. 1

56--64

EN

Physical activity (PA) is a well-known, simple and effective preventive and therapeutic intervention for low back pain (LBP). In spite of the growing interest in active lifestyles and its benefits, more needs to be known about the relationship between energy expenditure, body mass and lumbar-pelvic kinematics during the forward bending movement in a group of young asymptomatic people who met PA guidelines. Young people can be identified as a future risk group of civilisation diseases and lumbar-hip kinematics can be considered as a predictor of LBP occurrence. The aim of this study was to identify the association of gender, self-reported energy expenditure, body mass index, and lumbar-hip kinematics in young people. Methods: Sixty-four students at pre-employment stage participated in the study. They declared moderate-to-high PA and activity-induced energy expenditure (AEE) was self-reported. Kinematic data of the lumbar spine, pelvis and hip were collected during forward bending using a 3D motion ca pture system. Results: Sex was found to be associated with pelvis ( = –0.38 p = 0.002) and lumbar mobility ( = 0.49, p < 0.001) during forward bending and BMI was related only to lumbar mobility ( = –0.41, p = 0.001). Recreation AEE significantly predicted hip flexion mobility ( = 0.38, p = 0.002). Conclusions: This study showed that among a sample of physically active young people, BMI, self-reported AEE and sex can partially predict lumbar-hip kinematics during trunk flexion. Recreational PA can be regarded as improving hip mobility and thus making forward bending more effective and less prone to injury.

8

Recognition of human gait based on ground reaction forces and combined data from two gait laboratories

51%

Derlatka M. , Skublewska-Paszkowska M. , Powroźnik P. , Smołka J. , Łukasik E. , Borysiewicz A. , Borkowski P. , Czerwiński D.

Acta Mechanica et Automatica

|

2024

|

tom Vol. 18, no. 2

361--366

EN

In a world in which biometric systems are used more and more often within our surroundings while the number of publications related to this topic grows, the issue of access to databases containing information that can be used by creators of such systems becomes important. These types of databases, compiled as a result of research conducted by leading centres, are made available to people who are interested in them. However, the potential combination of data from different centres may be problematic. The aim of the present work is the verification of whether the utilisation of the same research procedure in studies carried out on research groups having similar characteristics but at two different centres will result in databases that may be used to recognise a person based on Ground Reaction Forces (GRF). Studies conducted for the needs of this paper were performed at the Bialystok University of Technology (BUT) and Lublin University of Technology (LUT). In all, the study sample consisted of 366 people allowing the recording of 6,198 human gait cycles. Based on obtained GRF data, a set of features describing human gait was compiled which was then used to test a system’s ability to identify a person on its basis. The obtained percentage of correct identifications, 99.46% for BUT, 100% for LUT and 99.5% for a mixed set of data demonstrates a very high quality of features and algorithms utilised for classification. A more detailed analysis of erroneous classifications has shown that mistakes occur most often between people who were tested at the same laboratory. Completed statistical analysis of select attributes revealed that there are statistically significant differences between values attained at different laboratories.