A Survey of Visual Commonsense Generation Task

Abstract

The task of extracting commonsense from data has received a lot of attention, and various research on commonsense tasks is conducted. This paper provides an overview of the visual commonsense generation(VCG) tasks using visual and language data. First, we will describe the differences between VCG tasks and other tasks that use commonsense such as reasoning, question, answering, and etc. Next, we will look at what datasets are available and what models are being used in relation to the VCG task. Finally, we will explain the limitations that are in the VCG task studied so far and will discuss future research directions in VCG task. This survey will provide an introduction to VCG task and a guide to practitioners.