dc.contributor.author | Βασίλη, Κωνσταντίνος | el |
dc.contributor.author | Vasili, Konstantinos | en |
dc.date.accessioned | 2019-03-29T08:19:26Z | |
dc.date.available | 2019-03-29T08:19:26Z | |
dc.date.issued | 2019-03-29 | |
dc.identifier.uri | https://dspace.lib.ntua.gr/xmlui/handle/123456789/48535 | |
dc.identifier.uri | http://dx.doi.org/10.26240/heal.ntua.15630 | |
dc.rights | Αναφορά Δημιουργού-Μη Εμπορική Χρήση 3.0 Ελλάδα | * |
dc.rights.uri | http://creativecommons.org/licenses/by-nc/3.0/gr/ | * |
dc.subject | Όραση υπολογιστών | el |
dc.subject | Νευρωνικά δίκτυα | el |
dc.subject | Ταξινόμηση εικόνας | el |
dc.subject | Βίντεο | el |
dc.subject | Computer vision | en |
dc.subject | Neural networks | en |
dc.subject | Object detection | en |
dc.subject | Image classification | en |
dc.subject | Video data | en |
dc.title | Αναγνώριση αντικειμένων σε εναέρια υψηλής ανάλυσης δεδομένα βίντεο με συνελικτικά νευρωνικά δίκτυα | el |
dc.title | Object detection from high resolution aerial video data with CNNs | en |
heal.type | bachelorThesis | |
heal.generalDescription | χρήση μεθόδων deep learning για αναγνώριση μικρών αντικειμένων και αξιολόγηση των μεθόδων με αυτόματη εξαγωγή ποσοτικών δεικτών αξιολόγησης | el |
heal.classification | computer vision | el |
heal.language | el | |
heal.access | campus | |
heal.recordProvider | ntua | el |
heal.publicationDate | 2018-10-10 | |
heal.abstract | Αναγνώριση αντικειμένων σε δεδομένα βίντεο τα οποία έχουν ληφθεί με εναέρια μέσα με τη χρήση συνελικτικών νευρωνικών δικτύων | el |
heal.abstract | As technology and the capabilities of UAVs continually increa se , data analysis and automated information extraction is imperative for a variety of geospatial, environmental and other app lications. In this diploma thesis, the most well - known deep learning methods for object detection from aerial means were studied, applied and evaluated . The data used consists of relatively high - resolution video s from the Stanford Drone Dataset collection, which includes a large number of different scenes in the university's campus. After some initial experiments, two recent techniques which had achieved high detection accuracy in well - known images collections for Computer Vision were selected . In particula r, the 'Faster R - CNN' and 'YoLo' method s for detecting and identifying objects in the 'Stanford Drone Dataset' w ere selected, studied and programmatically fine tuned. An algorithm in Python was developed, one for each method, for automatically extracting q uality indices i n order to compare the results . Βased on the results of the qualitative and quantitative evaluation , the 'Faster R - CNN' method was selected for further experimentation and analysis. In particular, a new set of experiments, fine - tuning and evaluation was carried out in order to determin e the possibility of generalizing the efficiency of the method over a large data range with significant variance in brightness, spatial resolution and detection difficulties | en |
heal.advisorName | Καράντζαλος, Κωνσταντίνος | el |
heal.committeeMemberName | Αργιαλάς, Δημήτριος | el |
heal.committeeMemberName | Γεωργόπουλος, Ανδρέας | el |
heal.academicPublisher | Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Αγρονόμων και Τοπογράφων Μηχανικών. Τομέας Τοπογραφίας. Εργαστήριο Τηλεπισκόπησης | el |
heal.academicPublisherID | ntua | |
heal.numberOfPages | 100 σ. | |
heal.fullTextAvailability | true |
Οι παρακάτω άδειες σχετίζονται με αυτό το τεκμήριο: