3D Gaussian splatting demonstrates significant advantages in high-fidelity rendering for offline reconstruction of objects or scenes. However, its integration with existing simultaneous localization and mapping (SLAM) systems, especially in monocular scenarios, still suffers from limited localization accuracy and poor rendering quality. In this paper, we propose a new monocular 3D Gaussian splatting SLAM, which achieves high fidelity online 3D Gaussian splatting based reconstruction given monocular video input with significance-guided pruning. Our key idea is to maintain structural compactness when optimizing the 3D Gaussians, by adaptively pruning them using a global significance evaluation based on multi-dimensional cues such as visibility, opacity, and volume coefficient. Specifically, we use a frame-to-model pipeline that jointly optimizes camera poses and 3D Gaussians within a sliding-window framework, ensuring a globally consistent 3D Gaussian representation for high-fidelity rendering with significance-guided pruning. Furthermore, an online monocular depth estimation model is incorporated to extract depth priors from input images, to effectively initialize the 3D Gaussian attributes for better camera tracking. Extensive experiments on Replica and TUM datasets demonstrate that our approach substantially improves both tracking performance and rendering fidelity, and thus provides state-of-the-art results.

Inspired by the rapid progress of generative AI techniques, there have been huge advances made for the 3D (three-dimensional) reconstruction community, which promoted the traditional 3D reconstruction framework from deep implicit 3D reconstruction to generative 3D reconstruction, achieving more robust and expansive 3D reconstruction results with the help of generative AI models. Meanwhile, there is still a lack of corresponding review articles to provide a comprehensive analysis of recent advances from the perspective of 3D reconstruction. In response, this paper gives a comprehensive review for the generative 3D reconstruction approaches, especially on the recent advances made from the computer graphics and vision communities. Firstly, this paper mainly divides the recent generative 3D reconstruction approaches into four categories, including generative structure-from-motion/multiview-sterero (SfM/MVS), generative adversarial networks (GAN) based 3D reconstruction, diffusion-based 3D reconstruction, and cross-modal 3D reconstruction, which cover most generative-model aided 3D reconstruction work with a comprehensive review and analysis. Thereafter, some representative applications inspired by the generative 3D reconstruction including dynamic human avatars, 3D interactive editing, and autonomous driving are also reviewed. Besides, some major datasets widely used for the generative 3D reconstruction approaches are included. Finally, this paper makes a discussion of the potential future work in further improving the quality of generative 3D reconstruction, towards better and more intelligent 3D reconstruction and generation.