0x02A

9 ms·

1. 1.▲
   by 0x02A 3y ago

   Oh I see. That's a pretty cool use case. Not sure how GS would perform on non-photorealistic scenarios, but certainly worth a try.
2. 2.▲
   by 0x02A 3y ago

   Does colmap fail on your dataset?
3. 3.▲
   by 0x02A 3y ago

   The original paper doesn't work well with few-shot learning. I'm assuming that there is only one camera angle for each pre-rendered background. For single image to 3D, check out DreamGaussian. [1] [1] https://arxiv.org&
4. 4.▲
   by 0x02A 3y ago

   That's super interesting. Are you trying to do single-view image to 3D?
5. 5.▲
   by 0x02A 3y ago

   Yeah, I gave it a try and timings seems to be very wrong. I'll fix that soon. I haven't tried benchmarking SPIR-V shaders on macOS. Since they're translated into Metal shaders anyways, it should be possible theoretically. Als
6. 6.▲
   by 0x02A 3y ago

   Thanks for trying it out! I haven't had the opportunity to benchmark this on an Intel Macbook. Were you able to see which kernel takes the most time? There should be a performance graph if you have the GUI enabled. For my Apple Silicon
7. 7.▲
   by 0x02A 3y ago

   Hey HN, this project started out when I was exploring rendering radiance fields in real-time for standalone VR and AR headsets. I was frustrated by the lack of performant, yet non-CUDA, implementations. Also, this would be a good excuse to
8. 8.▲
   Show HN: 3DGS.cpp – performant, cross platform Gaussian Splatting with Vulkan(github.com)

   50 points by 0x02A 3y ago|20 comments