Automatic image captioning for web accessibility as an Encoder-Decoder problem

Ramirez Garcia, Victor

Automatic image captioning for web accessibility as an Encoder-Decoder problem

License

Citation

Ramirez Garcia, V. Automatic image captioning for web accessibility as an Encoder-Decoder problem. 2020. handle: http://hdl.handle.net/10230/46067

This citation was generated automatically.

Abstract

Automatic Image Description is one of the most signi cant problems in Arti cial Intelligence, which consists of text synthesis for describing visual scenes. One of its main applications is to assist blind people in accessing information by using screen readers. In this project, we will study di erent methodologies for the generation of descriptions, posing the task as a problem of text synthesis through an Encoder-Decoder architecture, while also developing a model that can accomplish it. Finally, we will implement a series of improvements (changes in the structure or the use of visual attention mechanisms) to the previous model. We will evaluate the results, showing that it is possible to imitate those of the State-of-the-Art.