The field of 3D Gaussian Splatting is rapidly evolving, with a focus on improving the accuracy and efficiency of 3D reconstruction and novel view synthesis. Recent developments have addressed challenges such as sparse-view scenarios, multi-appearance conditions, and large-scale scene rendering. Notable advancements include the use of geometric priors, structure-from-motion points, and voxel-aligned prediction to enhance 3D consistency and reduce overfitting. Additionally, innovative applications of 3D Gaussian Splatting have emerged, including contactless fingerprint recognition, deformable surgical navigation, and aerial-ground image feature matching.

Noteworthy papers include MS-GS, which achieves photorealistic renderings under challenging sparse-view and multi-appearance conditions, and GS-Scale, which enables large-scale 3D Gaussian Splatting training on consumer-grade GPUs. VolSplat is also notable for its voxel-aligned prediction paradigm, which overcomes limitations of traditional pixel-aligned Gaussian prediction. Other notable papers include FixingGS, WaveletGaussian, and PU-Gaussian, which propose innovative methods for enhancing 3D Gaussian Splatting, such as training-free score distillation, wavelet-domain diffusion, and point cloud upsampling using 3D Gaussian representation.