LightSpeed: Light and Fast Neural Light Fields on Mobile Devices
Aarush Gupta, Junli Cao, Chaoyang Wang, Ju Hu, Sergey Tulyakov, Jian, Ren, L\'aszl\'o A Jeni

TL;DR
LightSpeed introduces a neural light field method using the light slab representation, enabling real-time, high-quality view synthesis on mobile devices with faster training and rendering compared to previous approaches.
Contribution
The paper proposes using the light slab representation for neural light fields, improving efficiency and extending applicability to non-frontal scenes on mobile devices.
Findings
Achieves real-time rendering on mobile devices.
Outperforms previous light field methods in quality and speed.
Extends to non-frontal views with a divide-and-conquer approach.
Abstract
Real-time novel-view image synthesis on mobile devices is prohibitive due to the limited computational power and storage. Using volumetric rendering methods, such as NeRF and its derivatives, on mobile devices is not suitable due to the high computational cost of volumetric rendering. On the other hand, recent advances in neural light field representations have shown promising real-time view synthesis results on mobile devices. Neural light field methods learn a direct mapping from a ray representation to the pixel color. The current choice of ray representation is either stratified ray sampling or Plucker coordinates, overlooking the classic light slab (two-plane) representation, the preferred representation to interpolate between light field views. In this work, we find that using the light slab representation is an efficient representation for learning a neural light field. More…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
Taxonomy
TopicsAdvanced Vision and Imaging · Robotics and Sensor-Based Localization · Computer Graphics and Visualization Techniques
