A New Framework for Robots that Can Draw Based on Labeled Examples
Keywords
No Thumbnail Available
Authors
Issue Date
2014-07-15
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This Master thesis describes a new framework, called NaoPLEx, for robots that can draw based on labeled visual examples. In the framework has been created an interaction in which a user teaches to the robot how to perform drawings. The interaction follows same the paradigm of teacher-pupil at elementary schools. The teacher (the user) draws something on the whiteboard and names it. Consequently the pupil (the robot) observes the drawing extracting features of interest from the drawing in order to re-create it as accurate as possible. In a second instance, the user asks to the robot to draw by itself a drawing already taught and the robot is able to do it accurately, drawing its own version of the example. Thus, in the experiments, the user has drawn examples on the whiteboard and has labeled them verbally. The robot has performed its own version of the drawings and the output drawings have been evaluated by human observers through a survey. The human observers rated the drawings as an accurate re-creation of the input. Moreover, the outputs have been given as input to the model again in order to rate the robustness of the model. It has came out that the model is not able to process the robot’s output due to heavy noises in the pictures. An Aldebaran NAO robot has been used during the project.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen