Automated Image Captioning: Exploring the Potential of Microsoft Computer Vision for English and Spanish

Diplôme Master Automated Image Captioning: Exploring the Potential of Microsoft Computer Vision for English and Spanish eng 2019 author Martinez Gutierrez Maria Fernanda 832503 director De Wilde Max 868907 With the rise of deep learning, reflected by the creation of architectures such as Convolutional Neural Networks (CNNs), researchers are becoming increasingly interested in the utility and relevance of machines that can properly generate information about images in different languages. This thesis focuses on Microsoft Azure's Computer Vision API, specifically its functionality for image description in both English and Spanish applied to a corpus of flora pictures. To assess the accuracy of the API's captions, a combination of human and machine evaluation was used. Although the initial hypothesis was that the CNNs of the API were robust enough to generate pertinent captions in both languages, the evaluations seemed to indicate that the technology is not yet mature enough to accomplish this task. This exploratory study therefore serves as a reflection on the use of automated image captioning for multilingual purposes and of the potential and limits of this technology. Automatic Image Captioning Deep Learning Convolutional Neural Networks CNNs Computer Vision Microsoft Azure Flora Machine Learning Machine-generated captions Dewey 410.2 Maîtrise universitaire en traitement informatique multilingue 112 Département de traitement informatique multilingue 92 false 62372665-f27b-457a-a097-b3a2d4dc4d78 Master thesis Public Thesis_Maria_MATIM.pdf 6806063 application/pdf openAccess Faculté de traduction et d'interprétation / Département de traitement informatique multilingue

Title	Published in	Access level	OA Policy	Year	Views	Downloads

Automated Image Captioning: Exploring the Potential of Microsoft Computer Vision for English and Spanish				2019	470	1,059

Archive ouverte UNIGE

Martinez Gutierrez, Maria Fernanda