Detection and Tracking of Objects on Water Surface
Detekce a sledování objektů na vodní hladině
Authors
Supervisors
Reviewers
Editors
Other contributors
Journal Title
Journal ISSN
Volume Title
Publisher
České vysoké učení technické v Praze
Czech Technical University in Prague
Czech Technical University in Prague
Date
Abstract
Hlavným cieľom tejto diplomovej práce bolo navrhnúť a implementovať systém, ktorý dokáže detekovať, klasifikovať a sledovať plávajúci odpad na vodnej hladine. Na detekciu a klasifikáciu plávajúceho odpadu boli navrhnuté a implementované viaceré metódy, ktorých základ bol v hlbokom učení. Navrhnutá a implementovaná bola rýchla a výpočetne nenáročná metóda, ktorá zvláda sledovať viacero objektov na vodnej hladine. Pre natrénovanie a vyhodnotenie modelov so základom v hlbokom učení bol nazbieraný dataset. Dataset sa využil aj na vytvorenie videí, ktoré boli neskôr použité k vyhodnoteniu sledovacej metódy. V závere našej práce prinášame prehľad o prevedených experimentoch a diskusiu o ich výsledkoch, na základe ktorých sme navrhli detekčný model, ktorý je vhodný na použitie so sledovacou metódou pri implementácii na reálny dron.
The main goal of this diploma thesis was to design and implement a system that can detect, classify and track the floating debris on the water surface. Methods based on deep learning were proposed and implemented for the detection and classification of floating debris. A fast and computationally not demanding method, which was able to track multiple objects on the water surface, was proposed and implemented. Dataset was collected for the training and evaluation of the deep learning models. The dataset was also used to create videos for the evaluation of the tracking method. At the end of the thesis, we present and discuss the experiments and results. We proposed a detection model, suitable for implementation with the multi-object tracking method on the real hardware of the UAV.
The main goal of this diploma thesis was to design and implement a system that can detect, classify and track the floating debris on the water surface. Methods based on deep learning were proposed and implemented for the detection and classification of floating debris. A fast and computationally not demanding method, which was able to track multiple objects on the water surface, was proposed and implemented. Dataset was collected for the training and evaluation of the deep learning models. The dataset was also used to create videos for the evaluation of the tracking method. At the end of the thesis, we present and discuss the experiments and results. We proposed a detection model, suitable for implementation with the multi-object tracking method on the real hardware of the UAV.