Μεταπτυχιακές Εργασίες

Μόνιμο URI για αυτήν τη συλλογήhttps://pyxida.aueb.gr/handle/123456789/7

Περιήγηση

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω

Τώρα δείχνει 1 - 1 από 1

Crop and weed detection using tensorflow object detection API
(01/01/2019) Alexis, Konstantinos; Athens University of Economics and Business, Department of Informatics; Arkoumanis, Konstantinos; Kotidis, Υannis
The purpose of this thesis is to study a specific object detection task, namely to develop an end-to-end pipeline, able to detect crop and weed instances in a field imagery dataset. Tensorflow Object Detection API was utilized, which has democratized the object detection field by making research code developed at Google available. Object detection key concepts along with two state-of-art approaches, commonly referred in the relevant bibliography, the Single Shot Detector and the Faster R-CNN, are presented. Afterwards, we demonstrate the steps to create an image dataset ready to be used in the API. We then present the experimental results, providing some remarks about the whole task. We find out that the SSD models seem to be faster, while Faster R-CNN tend to achieve a higher performance, making the choice of the best approach to be a matter of each specific application needs.