Visual Question Answering System for Indian Regional Languages
Dhiraj Amin, Sharvari S. Govilkar, Sagar Kulkarni · 2022
Questions are asked over the image within the Visual question answering system (VQAS). VQAS takes input as a user's natural language question and relative image. VQAS are mostly trained with a huge dataset with image, question and its correct answer. Many datasets like VQA, Visual7W and others have questions and answers over the image in English but there is scarcity of resources for many regional languages like Hindi and Marathi. In this paper we investigate different methods over the Hindi and Marathi version of the easy-VQA dataset.