3D Reconstruction and Estimation from Single-view 2D Image by Deep Learning – A Survey

3D Object Reconstruction from a single-view 2D image has become a promising research field. However, it remains a crucial and unsolved core issue in AI and Computer Vision research. Many scholars think it is the future of Artificial Intelligence and is well deserved at the irreplaceable heart of future AI research. In this paper, the research history of 3D object reconstruction is introduced, and the current state-of-the-art research methods and most novel results are investigated and discussed. A prediction for the best research methods and reconstruction model for this field is made. This paper also provided the essential clues and trends in 2D images or scenes inverse to 3D sceneries by 3D reconstruction. This interdisciplinary research area requires the researchers to have rich knowledge in but not limited to Computer graphics (Such as OpenGL), deep learning, Computer vision (like OpenCV), and the neurocognitive logics and principles of the cerebral cortex.