In this paper an unknown objects drawing model is proposed. Here we describe a technique that let us build a visual model of a word through images retrieved from Internet, enabling to learn any object at any time. This process is done without any prior knowledge of the objects appearance. However this information must be filtered in order to get the most meaningful image according to the keyword and allowing making the visual relation between words and images as much unsupervised as it could be possible like humans understanding. For this purpose Pyramid of Histogram of Oriented Gradients (PHOG) feature extraction, K-means clustering and color segmentation is done, and then the final image is drawn as an application of the learning process. The proposed model is implemented in a robot platform and some experiments are carried out to evaluate the accuracy of this algorithm.
|Original language||American English|
|Number of pages||130|
|Journal||Revista Facultad de Ingenieria|
|State||Published - 1 Dec 2011|