SLS4D: Sparse Latent Space for 4D Novel View Synthesis
This work addresses efficiency and scalability issues in dynamic 3D scene reconstruction for applications like virtual reality and robotics, representing an incremental improvement over existing dynamic NeRF methods.
The paper tackles the problem of high parameter costs and limited global dynamics capture in dynamic neural radiance fields (NeRFs) for 4D novel view synthesis by proposing SLS4D, a sparse latent space representation that achieves state-of-the-art synthesis with only about 6% of the parameters of recent methods.
Neural radiance field (NeRF) has achieved great success in novel view synthesis and 3D representation for static scenarios. Existing dynamic NeRFs usually exploit a locally dense grid to fit the deformation field; however, they fail to capture the global dynamics and concomitantly yield models of heavy parameters. We observe that the 4D space is inherently sparse. Firstly, the deformation field is sparse in spatial but dense in temporal due to the continuity of of motion. Secondly, the radiance field is only valid on the surface of the underlying scene, usually occupying a small fraction of the whole space. We thus propose to represent the 4D scene using a learnable sparse latent space, a.k.a. SLS4D. Specifically, SLS4D first uses dense learnable time slot features to depict the temporal space, from which the deformation field is fitted with linear multi-layer perceptions (MLP) to predict the displacement of a 3D position at any time. It then learns the spatial features of a 3D position using another sparse latent space. This is achieved by learning the adaptive weights of each latent code with the attention mechanism. Extensive experiments demonstrate the effectiveness of our SLS4D: it achieves the best 4D novel view synthesis using only about $6\%$ parameters of the most recent work.