Review of Recent Datasets Used in Image Captioning Models

Automatic image captioning is a complex research area whose objective is to generate descriptions in natural language from a query image. It has significantly advanced with deep learning with applications in humancomputer interaction, medical image captioning, traffic analysis, industrial quality control, and assistive systems for the visually impaired. Image captioning task includes visual features extraction and generate syntactically and semantically correct descriptions of the image. Deep learning techniques are proficient in handling the multifaceted relationships among image objects. This work explores 14 popular image captioning datasets used in the evaluation of deep learning models for automatic image captioning. The ultimate goal of this work is to provide better understanding of image captioning datasets which may benefit the related research communities.