Overview of Scene Graph Generation Approaches in Computer Vision

Scene graph generation, the task of identifying objects in an image and their relationships, has numerous applications in computer vision, including image captioning and object detection. While deep learning has shown promise for this task, hyperparameter tuning remains a challenge. This paper considers the background and overviews for understanding the visual scene, scene graph and Scene Graph Generation. We discuss the visual scene related research area and scene graph concept. Also we follow the chronological and methodological approaches for the scene graph generation. Then we go through the neural network and deep learning approaches to scene graph generation (e.g. human-object/human-human interaction).