Image Captioning in Turkish Language

Abstract

Image captioning is one of the everlasting challenging tasks in the field of artificial intelligence that requires computer vision and natural language processing. Plenty of salient works have been proposed throughout the time for English language however, the number of studies in Turkish language is still too limited. This paper couples an encoder CNN-the component that is responsible for extracting the features of the given images-, with a decoder RNN -the component that is responsible for generating captions using the given inputs-to generate Turkish captions within human gold-standards. We conducted the experiments using the most common evaluation metrics such as BLEU, METEOR, ROUGE and CIDEr. Results show that the performance of the proposed model is satisfactory in both qualitative and quantitatively evaluations. A Web App is already deployed to allow volunteers to contribute to the improvements of the Turkish captioned dataset.