Πλοήγηση ανά Συγγραφέας "Kyprakis, Ioannis"

Τώρα δείχνει 1 - 2 of 2
  • Μικρογραφία εικόνας
    Τεκμήριο
    Calculation of vegetation indicators in crops through optical methods and IoT technologies.
    (ΕΛ.ΜΕ.ΠΑ., ΣΧΟΛΗ ΜΗΧΑΝΙΚΩΝ (ΣΜΗΧ), Τμήμα Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών, 2022-10-14) Kyprakis, Ioannis; Κυπράκης, Ιωάννης
    Calculation of vegetation indicators is very important to understand a plant’s health, and how the plant behaves under certain environmental conditions. Those calculations are extremely helpful to agronomists and farmers and have applications in various fields such as agronomy, biology, botany, and many others. Many devices are made for this purpose, but most come with an excessive cost, low precision, or are too complex to use. This thesis proposes a precise, low-cost, easy-to-use device that calculates crop vegetation indicators through non-destructive, contactless optical methods and IoT technologies. The optical methods we use to calculate the indices are based on light reflection. The light reflected from the leaf’s surface will be measured. Some of the vegetation indices, that we will calculate are Simple Ratio, Normalized Difference Vegetation Index (NDVI), NDVIg, NDVI¬¬¬b), Infrared Percentage Vegetation Index (IPVI), and Enhanced Vegetation Index (EVI). We will evaluate with experiments the precision of our device. We will use LEDs that emit at certain wavelengths (465nm, 535nm, 630nm, and 840nm) and measure the reflectance from the surface of the leaves. The comparison revealed similar performance, demonstrating a strong correlation with the HR2000+ spectrometer (R2 = 0.92-0.97), proving the device's high potential for precise plant stress measurements. The hardware implementation consists of an Arduino Mega, sensors, modules, and other electronic components. We will also use I2C and SPI protocols to achieve communication between our microcontroller and the modules. The device will measure the vegetation indicators and then the values will be stored on an external SD card. The software part of the device was implemented using Arduino IDE v1.8.19.
  • Μικρογραφία εικόνας
    Τεκμήριο
    Identification of Parkinson’s disease facial symptoms by utilizing deep learning techniques.
    (ΕΛΜΕΠΑ, Σχολή Μηχανικών (ΣΜΗΧ), ΠΜΣ Μηχανικών Πληροφορικής, 2025-01-21) Kyprakis, Ioannis; Κυπράκης, Ιωάννης; Tsiknakis, Emmanouil; Τσικνάκης, Εμμανουήλ
    Parkinson’s Disease (PD) is a progressive neurological disorder characterized by motor and non-motor symptoms, including hypomimia, a significant reduction in facial expressiveness. Hypomimia, in conjunction with other symptoms, complicates the accurate assessment of depression in PD patients. This research presents a novel approach for estimating depression in PD patients by analyzing facial video recordings using deep learning techniques. The study leverages two datasets: a comprehensive video dataset comprising 1,875 facial recordings from 173 patients, and a detailed clinical dataset consisting of 140 patients and 18 features, including critical patient information such as UPDRS III scores and LEDD. Our results demonstrate that the 3D-CNN-LSTM model with attention outperforms other models, achieving an accuracy of 82% in binary classification and 71.17% in multiclass classification tasks. Furthermore, machine learning models applied to clinical data to reinforce the findings, highlighting key features such as UPDRS III scores and LEDD as significant predictors of depression. This research establishes a new benchmark for the automated detection of depression in PD patients, emphasizing the potential of integrating facial video analysis with clinical data for more accurate and early diagnosis.