Image synthesis designed for machine learning applications provides the means to efficiently generate large quantities of training data while controlling the generation process to provide the best distribution and content variety. With the demands of deep learning applications, synthetic data has the potential of becoming a vital component in the training pipeline. Since 2010, a wide variety of training data generation methods has been demonstrated. The potential of future development calls to bring these together for comparison and categorization.
This survey provides a comprehensive list of the existing image synthesis methods for visual machine learning. These are categorized in the context of image generation, using a taxonomy based on modeling and rendering, while a classification is also made concerning the computer vision applications they are used. We focus on the computer graphics aspects of the methods, to promote future image generation for machine learning. This page along with the corresponding report serve as a reference, targeting both groups of the applications and data development sides.
The table below contains image generation techniques and synthetic datasets that have been derived from some of the methods discussed in the survey. Not all of the image synthesis methods that are included in the above pipeline are accompanied by a standalone, ready-to-use synthetic dataset. Whenever available we link to the generated training set, or to the generation pipeline code.
The list contains dataset details, like size, resolution and originally provided ground truth labels. The following does not aim to be a limited, static representation and therefore if you would like more image synthesis methods and/or synthetic datasets for visual machine learning to be included and featured, please contact us with the relevant information.