Deep learning techniques have attracted many researchers in computer vision field to solve computer vision problems such as image segmentation and object recognition. This success also led to the implementation of deep learning techniques in 3D reconstruction. 3D reconstruction itself is a classical problem in computer vision that has been approached by many techniques. However, deep learning techniques for 3D reconstruction are still in the early phase, whereas the opportunity of the techniques is large. Hence, to improve the performance of such approaches, it is important to study the research and applications of the approaches through literature review. This paper reviews deep learning-based methods in 3D reconstruction from single or multiple images. The research scope includes single or multiple image sources but excludes RGB-D type input. Several methods and their significance are discussed, also some challenges and research opportunities are proposed for further research directions.