The field of novel view synthesis is rapidly advancing, with a focus on improving the efficiency and quality of 3D Gaussian Splatting (3DGS) methods. Recent developments have explored new approaches to enhance the performance of 3DGS, including disentangled 4D Gaussian Splatting, augmented 3D Gaussian Splatting, and neural pruning for 3D scene reconstruction. These innovations have led to significant improvements in rendering speed, storage requirements, and image quality. Notably, the introduction of spatial condition-based prediction and instance-aware hyper prior have enabled effective compression of 3DGS models, reducing storage costs and transmission times. Furthermore, the application of Gaussian Splatting to new domains, such as imaging sonar, has demonstrated promising results. Overall, the field is moving towards more efficient, compact, and high-quality 3D representations. Noteworthy papers include:

• Disentangled 4D Gaussian Splatting, which achieves an unprecedented average rendering speed of 343 FPS.
• NeuralGS, which reduces the model size of 3DGS by 45 times without harming visual quality.
• Enhancing 3D Gaussian Splatting Compression, which achieves a 24.42 percent bit rate savings compared to the state-of-the-art compression method.