Abstract—The fast progress of Three-Dimensional (3D) reconstruction has led to the emergence of advanced Deep Learning (DL) approaches and techniques. Leveraging the technology of computers to produce realistic three-dimensional representations of objects has grown into an essential component of extensive study in a variety of domains. This review article investigates the cutting-edge methodologies, difficulties, and potential in this research field. The state-of-art study follows the development of Deep learning techniques with graphics expertise, which strengthens the requirement for good efficacy with better performance of 3D reconstruction. The research work begins by discussing classic strategies for 3D reconstruction with active and passive techniques that emphasizes their limitations with the need for cutting-edge practices. The various types of neural network architectures employed, like Convolutional Neural Networks (CNNs), autoencoders, and Generative Adversarial Networks (GANs) are explored with auxiliary information. This review aims to provide researchers and practitioners with a thorough understanding of the advancements, problems, and prospects in image-based 3D reconstruction while opting for the progressions in Deep Learning. Further, this research study presents the development in Neural Radiance Fields (NeRF) which is revolutionizing image-based rendering for efficient 3D reconstructions.
 
Keywords—Three-Dimensional (3D) reconstruction, Deep Learning (DL), Convolutional Neural Network (CNN), Generative Adversarial Network (GAN), Neural Radiance Fields (NeRF)


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).