In plain words: A survey of 25 recent studies and 6 datasets on systems that answer questions about pictures, comparing how each one works and scores. It shows which approaches lead today and the mistakes they still make most often.

Abstract · Visual Question Answering: A Survey on Techniques and Common Trends in Recent Literature

Visual Question Answering (VQA) is an emerging area of interest for researches, being a recent problem in natural language processing and image prediction. In this area, an algorithm needs to answer questions about certain images. As of the writing of this survey, 25 recent studies were analyzed. Besides, 6 datasets were analyzed and provided their link to download. In this work, several recent pieces of research in this area were investigated and a deeper analysis and comparison among them were provided, including results, the state-of-the-art, common errors, and possible points of improvement for future researchers.

Ana Cláudia Akemi Matsuki de Faria, Felype de Castro Bastos, José Victor Nogueira Alves da Silva, Vitor Lopes Fabris, Valeska de Sousa Uchoa, Décio Gonçalves de Aguiar Neto, Claudio Filipi Goncalves dos Santos
arXiv:2305.11033 · cs.CV, cs.AI, cs.LG · submitted May 18, 2023 · updated Jun 2, 2023
abstract · pdf · html · 30 pages. arXiv admin note: text overlap with arXiv:2104.00926, arXiv:2110.02526, arXiv:2108.02059, arXiv:1908.01801 by other authors