The recent release of a groundbreaking technology has sparked considerable excitement across the tech community.
I am of course referring to the code release of Trilinear Point Splatting for Real Time Radiance Field Rendering, otherwise known as TRIPS. The engineer, Linus Franke, has released the code to TRIPS with a full MIT License. TRIPS stands out for several reasons. First, it significantly improves fidelity, going beyond the capabilities of 3D Gaussian Splatting. Additionally, it maintains the hallmark view-dependent features that make radiance fields so realistic. Most importantly, it achieves this while still ensuring real-time rendering, despite utilizing a small neural network.
https://www.youtube.com/watch?v=Nw4A1tIcErQ
Like the vast majority of radiance field methods, you will need an NVIDIA GPU to run TRIPS, but it seems to be fairly VRAM efficient, as Linus stated that it has been tested with a RTX2070, which comes with 8GB. It’s thrilling to see such a high-fidelity method made accessible to users with consumer-level graphics cards. As soon as I first read the paper I was excited to be able to give it a try and now it’s finally here (it’s only been a few weeks. I’m just impatient).
The installation process turned out to be smoother than anticipated. On a scale of 1 to OG Instant NGP, this is a breeze. The Github page does a good job of providing step by step instructions on how to install it on either Linux or Windows. I only ran into trouble installing CuDNN, but this article helped me.
In addition to the method itself, TRIPS also comes with the ADOP viewer. It’s a UI deviation from existing radiance field viewers, but I actually having the multiple windows visualized. It feels a bit more natural for me to compare the views while generating keyframes to render out.
Where I have run into trouble is converting COLMAP data into the correct format that ADOP can read, which is what TRIPS utilizes. The author, Linus, has outlined how to do so, but I have not yet had a chance to go through it for myself.
Meanwhile, in seemingly unrelated news, a well-known Cupertino-based tech giant also made headlines with a headset launch last Friday.
New to Gaussian Splatting? Start here
