>This site was suspended as it reached the limits of the Free plan
Anyhow the novel approach opened the route for building on top. A paper recently proposed dynamic captures, for real time animated Gaussian Splats.
It will go on there is so much to explore research-wise, optimisation like this innovation along is a large field of efforts.
That paper kicked off a rapid stream of a thousand papers by taking a photogrammetry-style workflow and producing better than photogrammetry results by reframing the process as gradient decent on differentiable point samples. This allowed the research to stand on the shoulders of all the work being put into deep learning tech.
Typical HN.
https://x.com/janusch_patas publishes a steady stream of it every day on X and at https://radiancefields.com/
The early 2000s splatting from point-based rendering is what George Drettakis and his students realized could be applied in this new NeRF domain.
Basically, all the reasons that point splats didn't work for regular surface rendering nearly 25 years ago (holes, inefficient, can't edit them like meshes) are less of an issue in a light field capture setup.
jauntywundrkind•7mo ago
Crazy good insight here: splatting is largely a search problem, and that can be offloaded to the cloud.
> Specifically, on the cloud side, we propose asynchronous level-of-detail search to identify the necessary Gaussians for the client. On the client side, we accelerate rendering via a lookup table-based rasterization.