Replying to @DuckbillStudio
That makes sense. And please keep originals. One day we could run SH4 on that scale :p
1
35
Replying to @DuckbillStudio
How are your 16M splats 1.4gb when my 8M splats are like 2gb teach me the secret please. Not SOG right?
1
99
Replying to @Yeq6X
The link huggingface.co/spaces/multim… doesnt seem to work for me, there's an error on every image :(
30
Replying to @Yeq6X
someone try a gaussian splat of this please, or help me figure it out, I can do all the SFM and splatting myself
1
23
Replying to @thesimaximalist @Alacritic_Super
SLAM DUNK is the new frontier too of benchmarks approach described here
7
Replying to @thesimaximalist @Alacritic_Super
Also LaMARia seems to be the benchmark not the new approach for SLAM thats performing this well
9
Replying to @YeheLiu
No idea what you mean. His end game is splats and he’s eschewing AI depth maps here in exchange for real ones from stereo.
How is photogrammetry related here?
48
Replying to @Bz370708245542B @Afc_news360
Thats europe no? They’re not allowed to carry pencils
1
8
925
If AI can modify its own weights at runtime, wouldn’t that allow it to tweak itself constantly based on tasks?
18
Replying to @zunnanxu
So just to give you an idea on where things are in this space on the user side, I frequently run 3000 photo datasets of 24mpx and running them for two days is sometimes needed (for 6000 or 7000 photos)
I’m definitely happy to wait if there’s a good reason
13
