Extensive Study on Color and Light Translation of 2D Images using Machine Learning Approaches

Abstract

For machine learning applications, digital image production provides for the efficient generation of huge volumes of training data while preserving control over the generation process to ensure the optimal content distribution and variation. Synthetic data has the potential to become an important element of the training pipeline as the demand for deep learning applications grows. Over the last decade, a broad range of strategies for producing training data have been presented. The collecting of these for comparison and categorization is required for future improvement. This study presents a complete list of available visual machine learning image synthesis approaches. In the context of 2D picture production, these are classed as light transfer and colour transfer. The focus is on the computational features of approaches for developing machine learning colour transfer between image-to-image translation in the future. Finally, the learning potential of each approach is assessed based on its reported quality and performance. The study is meant to serve as a complete reference for both data and application developers. This is a comprehensive list of all the methods and approaches discussed in this page.