Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
MrNeRF
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
3 ms
·
1.
▲
$1700 bounty delivers 2.4× speedup for CUDA Gaussian Splatting training
(github.com)
3 points
by
MrNeRF
1y ago
|
1 comments
2.
▲
by
MrNeRF
1y ago
This started as a $1700 bounty to optimize the CUDA Gaussian Splatting renderer. The PR delivers impressive performance improvements with a 2.4× speedup. Training time for 7 scenes from the mipnerf360 dataset dropped from 50 minutes to ~20
3.
▲
Optimzed Fused-SSIM
(github.com)
2 points
by
MrNeRF
2y ago
|
1 comments
4.
▲
by
MrNeRF
2y ago
I significantly improved the forward and backward pass, as well as the inference of SSIM. Optimized Fused-SSIM (Forward) 2.49ms (Backward) 2.68ms (Inference) 1.43ms Original Fused-SSIM (Forward) 3.66ms (Backward) 3.52ms (Inference) 2.59ms T
5.
▲
by
MrNeRF
3y ago
Open source software deserves celebration. Its impact on the world is profound and transformative. Recently, I began porting the 3D Gaussian splatting paper to C++ and CUDA with an aim to optimize the training speed of such models. This end