The field of 3D reconstruction and video quality assessment is rapidly evolving, with a focus on improving the accuracy and efficiency of these tasks. One of the key directions is the development of novel methods for reducing the computational complexity of 3D reconstruction, such as the use of active learning and hierarchical Gaussian splatting. Another area of research is the improvement of video quality assessment metrics, with a focus on developing metrics that can accurately capture the perceptual impact of distortions introduced by advanced rendering techniques. Additionally, there is a growing interest in the use of neural uncertainty maps for active view selection in 3D reconstruction, which can help to reduce the number of required viewpoints while maintaining reconstruction accuracy.

Noteworthy papers in this area include the introduction of a novel batch-mode active learning policy that reduces training data requirements by 75% while improving accuracy, and the development of a memory-efficient framework for high-resolution 3D reconstruction using hierarchical Gaussian splatting. The proposal of a full-reference video quality metric that outperforms existing metrics and the introduction of a novel VQA framework that incorporates free-energy-based self-repair are also significant contributions.