Compositional Scene Representation Learning via Reconstruction: A Survey

Visual scene representation learning is an important research problem in the field of computer vision. The performance of artificial intelligence systems on vision tasks could be improved if more suitable representations are learned for visual scenes. Complex visual scenes are composed of relatively simple visual concepts, and have the property of combinatorial explosion. Compared with directly representing the entire visual scene, extracting compositional scene representations can better cope with the diverse combinations of background and objects. Because compositional scene representations abstract the concept of objects, performing visual scene analysis and understanding based on these representations could be easier and more interpretable. Moreover, learning via reconstruction can greatly reduce the need for training data annotations. Therefore, reconstruction-based compositional scene representation learning has important research significance. In this survey, we first outline the current progress on this research topic, including development history and categorizations of existing methods from the perspectives of modeling of visual scenes and inference of scene representations; then provide benchmarks, including an open source toolbox to reproduce the benchmark experiments, of representative methods that consider the most extensively studied problem setting and form the foundation for other methods; and finally discuss the future directions of this research topic.