agreed on everything, and myself moving around and the sun changing positions invariably changes the data.
check out this splat, incredible detail when viewed from my dorm “window” and any other perspective is unintelligible
This is extremely neat! Took a while to figure out where the window is in the black void behind the camera. It was super fun crossing the street to see how things look from the other side!
I was wondering if shadowing, lens flare, vignetting, prismatic scattering/reflections, DOF, and auto white balancing might cause those weird little artefacts. . .
white balance was locked, otherwise i would agree yes.
BUT!!
i found a huge source of floaters. about 30 images including a poster have been tracked to the wrong place. i have two of the identical poster in my room and it seems to have confused the program. im currently going through by hand and finding these and removing them before i re-run all the data
Somehow this immediately made me think of 2 things to try:
This should be interesting because the coloring is structural and possesses strange properties (e.g., different color when viewed from different angles) that are not realizable by Gaussian splatting, so I’m super curious how the program handles it.
In particular, if the LED is close to the focal point, the virtual image should be very far back, and thus occupy the same physical location as some real objects in the background. This sort of optical phenomenon is again not realizable by a Gaussian splat, so I’m curious what happens.
It might be interesting to also just take a bunch of photos into the reflector of a flashlight, and see where the reflected image of the LED ends up.
I believe both of these can be achieved. reflections can be created like on the surface of my phone and desk, plus there’s a fantastic scan of a butterfly on the same website my scan is hosted. The butterfly wings are iridescent and you can see the colours shift and shimmer.
I will attempt this soon.
I would also like to take photos of a screen and advance the video by one frame every x° to then encode video into space
Thanks for the pointer! I saw this butterfly splat which does replicate the iridescence. Looks like it’s being done by having multiple thin layers of ellipsoids, each with a different color and orientation, so that from certain viewing angles some orientations are hidden while others are visible.
I’m curious what it would do with larger-scale optical phenomena, like the virtual image formed by a convex lens. Here there is more than just color information–there is a very specific virtual image with a well-defined shape and apparent location.
This is amazing @Q. Thanks for showing Ruckusmattster !
I will work on that today once I get a better version of my room posted
i havent yet found a spare reflector, i did find a cd though.
sadly im limited to using my phone camera while away from university, this introduces variable img settings, low resolution, mandatory image stabilisation, and a myriad of other issues with the processing applied. the splat is worse. but here it is
i will try again when i have access to a real camera
I only watched the first SuperSplat in this thread.
It’s very cool.
I don’t think I’ve seen anything like that before.
It seems to take an awful lot of work to produce, but you can’t argue with the end result. 
Thanks for doing this! The image quality is certainly not as good as the room splat for the reasons you stated, but it does demonstrate how the program handled the reflection+iridescence.
Similarly to the mirror, it created extra physical space under the CD (into the table) to realize the reflections. Here’s the view from under the CD:
Reminds me of the shimmer in the film Annihilation (2018).
im hoping thats only due to a lack of a second perspective, if the disc were upright it might not do that - of course i shouldve throught about that before the scan
ive been playing around with an old scan
Breedlove.. 
call your mom…
Cheers 
