Vision Language Models: Methods, Datasets and Training, and Bibliometric Analysis

Artificial intelligence (AI) has been applied in various domains and has revolutionized human life, particularly large models (e.g., large language models (LLMs) and vision language models (VLMs)). VLMs is an emerging AI method that learns multimodal tasks from images and texts and generates multiple forms of data such as text, images, and videos, which has been applied in many domains such as robotics and autonomous driving and is a recent significant trend. However, there are no papers that analyze the current studies of VLMs by bibliometrics. In this paper, we analyze the state-of-the-art open-source VLMs for guiding readers to choose their VLMs. Furthermore, we summarize the datasets for training VLMs and the training processing of VLMs. Finally, we develop a bibliometric method to analyze the current publications of VLMs, including publication number, aggregation type, top authors, top affiliations, and top venues, which is a significant contribution to providing data analysis of the current development of VLMs. This paper contributes to providing introductory material including methods, datasets and training, and bibliometric analysis.