I left Unity at start of 2022. So for this lazy Tuesday afternoon, figured I can share the rambles I wrote in my goodbye email. No big insights there, just old man reminiscing. And hey the text is already written, so that makes it easy to copy-pasta it into the blog:

It’s now exactly 16 years of me working at Unity. And as everyone knows, if you don’t leave after 16 years, you have to wait until 32 years for your next chance (look, I don’t make the rules). It’s time for me to log off and watch this thing from the outside.

So, yeah, bye! Thanks everyone for this amazing journey! ❤️

…

However! @alexmclean suggested I should write up some random bits of trivia from the old times. You know, like your grandpa is always telling stories from his youth, and you have to listen no matter whether you like it or not. So here goes!

Back in 2005, I was looking for a new job. Applied to several companies (NVIDIA, Lionhead, DICE, Rockstar) mostly to no response. But! Some Tim Sweeney from Epic Games reached out and offered me a job. That went all the way through to an on-site interview (which was my first trip to the US, score!), where I did not do well on the technical interview, and that was a “nope” from Epic. That was sad!

However, several months later this company I’ve never heard about (“Over the Edge Entertainment”), making a game engine I’ve never heard about (“Unity”), wrote to me with a subject “Working on the future of middleware”, and text like:

We want to talk to you about joining our team and engineer this revolution with us. You’ll lead the PC version, and get to define your own projects, doing the coolest tech this side of the Sun. Make a mass market tool that can change how games are made…. write code to change the world.

To which, as any sane person would do, I replied “cool, but no thanks, good luck”. The world is filled with random 20-somethings imagining they can pull off the next big thing. But then they invited me over to a game jam they were organizing. We made this game called “Pakimono”:

And so I started working at Unity on 2006 January 4th. I absolutely thought that after a year or two this whole thing will go down as another cute, but failed attempt at making yet another game engine. Because these were the odds! But hey, at least it will be fun while it lasts.

Back then “heck yeah!” (alternative phrasing for “Win Wire” of today) moments were… sometimes strange. Like here’s @joe being extremely excited about getting a car model for the website. Why would our website have a car? No one knows.

Typical office experience was slightly different from the offices we have these days (or, well, used to have before 2020). Here’s @david sleeping in the Copenhagen office mid-2006:

End of 2006 was an “oh shit” moment when Microsoft shipped “XNA Game Studio Express”, a free game development environment based on C#, aimed at indies and small developers. That totally could have killed us; a handful of people against the whole Microsoft behemoth. Luckily, they never really focused on it, and by 2010 XNA was dead. There’s probably a lesson there of some sorts.

In 2007 I convinced one of my buddies (@valdemar) to join Unity, to work on porting it to our first console platform! That was the Nintendo Wii, of course (in retrospect, not the greatest platform for everyone except Nintendo itself). But, this started a long journey of non-PC based platform support in Unity.

In 2008 I made this (crappy, first time) felt work and in retrospect, have correctly predicted what our stock symbol will be a dozen years later. If that does not make me a professional analyst, I don’t know what would.

On a more serious note, in 2008 we started to port Unity editor to Windows (it was Mac-only before then). And given that I was pretty much the only person using Windows to begin with, that meant I was doing the port, yay (we’re still suffering the consequences today, I’m sorry). Anyway, in February we had this, which isn’t much but it’s a start:

And by 2008 September we had the first internal build for testing, in all 15.4 megabytes download size glory:

Overall 2008 was quite an eventful year. For example, we opened the brand new Kaunas (Lithuania) office, which (I think) was the first development office outside of Copenhagen. It looked as “fancy” as this, since that’s what we could afford at the time:

Fun fact: ten years later some scenes of HBO Chernobyl were filmed in the same building.

Also in 2008, I remember us watching Apple’s keynote where they announced the App Store and the whole SDK for creating third-party applications. We were “ok so this is pretty much the same as a Mac, it’s gonna be an easy port, let’s do it!!!1”. Famous last words, eh. Anyhoo, I somehow convinced @rej and @mantas to join in making the Unity iPhone port. The very first version, in a very rough state, shipped towards the end of 2008. And probably have affected the whole industry for decades to come.

Spring of 2009 saw us ship Unity 2.5 with the Windows Editor, and increased our game developer market by 10x or so. @amir had a suggestion to use this on the main website, to really drive down the message:

By 2010 the company grew to enormous size, for example here’s a whole company meeting:

I found this in my 2010 “yearly performance review / feedback”. Cute! Also, some things don’t really change.

2011, we had another massive “oh shit this could be bad” moment, when Adobe decided to add 3D capabilities to Flash (“Stage3D” as it would be called). Our web browser plugin was still going strong, and Flash was still huge. This could have been really, really bad! So we did the only thing that seemed to make sense at the time, which was “let’s add Flash as a platform for Unity”. Which meant converting C# code into ActionScript, and adding support for Flash’s custom 3D graphics API and their own strange shader language. Anyway, it (kinda, mostly) worked: “That’s a completely standard shader, written in Cg using our ‘surface shaders’ thing, compiled into Direct3D9 assembly by Cg, parsed into bytecode by mojoshader & converted to Flash’s shader bytecode (AGAL) by, well, me.”

Anyway, by 2014 Flash, including all Stage3D stuff, was pretty much dead. There’s probably another lesson in there somewhere, maybe like “don’t be too afraid of the last attempt of a dying platform to make itself relevant”, or something.

Anyway, all that C# -> ActionScript conversion experience turned out to be quite useful when in 2014 Apple decided that all iOS apps must be 64-bit from now on. Scripting and mobile folks scrambled to get that working by creating C# -> C++ converter a.k.a. IL2CPP, since our existing scripting technology at the time (Mono) simply did not work on 64-bit iOS devices.

I was not involved in any of that, but in spring of 2014 we got an invitation for a secret project on-site at Apple. Two engineers were to be sent in there for a month, without being aware of what they would be doing. So me and @alexey went, and it turns out the secret project was Metal graphics API. Against all odds and much to everyone’s surprise, Apple was the first one to ship a “new/modern” graphics API (before Microsoft DX12 and Khronos Vulkan). A month at Apple was an… interesting experience. We did get Unity working on Metal in that month, but for the WWDC conference Apple decided to go with a keynote demo from Epic/Unreal instead. They were still friends back then, eh :)

Oh, here’s a funny thing I found from 2015, in an email from Joachim about a core team’s hack week:

The goal. How do we architect Unity so it can open a 100GB project and you start working in it, in less than 10 seconds. <…> To do this right I think we’ll have to soon form a 2-3 person asset pipeline team that can own this and get it to completion.

Yeah, maybe we should get onto that :)

One recent thing I’m fairly proud of, is that in 2019 we managed to create what used to be called the “Core Kaunas” team (now “Quality of Life” team – yeah we’re not great at names). A very small team of mostly junior people, doing “random improvements all over the place”. Day to day it does not sound much, but now when I read our 2019, 2020 and 2021 (ed: these were all links to internal Unity docs) summaries, eh it’s not too bad! Besides all the things we actually did, maybe we have influenced some other teams to also work on “quality of life” improvements. Rock on.

All in all, github (on our “main” code repository) says that my overall contribution over the years has been around minus one million lines of code, so 🎉. Why it thinks I’m #1 in the amount of code commits, I’ve no idea.

That’s it! Do good work, and take care of each other.

And that was it! Maybe next time I should write about what the heck I’ve been doing since I left. Or maybe something else.

Over the past month it seems like Gaussian Splatting (see my first post) is experiencing a Cambrian Gaussian explosion of new research. The seminal paper came out in July 2023, and starting about mid-November, it feels like every day there’s a new paper or two coming out, related to Gaussian Splatting in some way. @MrNeRF and @henrypearce4D maintain an excellent list of all things related to 3DGS, check out their Awesome 3D Gaussian Splatting Resources.

By no means an exhaustive list, just random selection of interesting bits:

Ecosystem and tooling

• PlayCanvas has released SuperSplat, an online splat editor (source), have added 3DGS support to the engine v1.67.0 and made their own splat data compression approach (blog post) that is somewhat based on my blog posts, yay.
• Luma AI has released Unreal Engine 5 plugin and a WebGL Three.js based viewing library: Luma WebGL Library.
• Spline has added gaussian splat support, with simple editing tools.
• Discussion about possible “universal splat format” among various open source tool authors.

Research

• GaussianEditor: Swift and Controllable 3D Editing with Gaussian Splatting applies ML-based editing and inpainting tools for gaussian splats.
• Relightable 3D Gaussian: Real-time Point Cloud Relighting with BRDF Decomposition and Ray Tracing attempts to make gaussians re-lightable by associating typical PBR material parameters (normal, BRDF properties), incident light information, etc.
• GS-IR: 3D Gaussian Splatting for Inverse Rendering extends 3DGS pipeline to derive surface normals, albedo and roughness, to achieve re-lighting of gaussian splats.
• LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction reduces gaussian splat data sizes by pruning and merging splats to reduce their count, makes spherical harmonics data smaller by reducing their order more cleverly than just “drop some coefficients”, and applies several quantization schemes to final data. This is clever stuff!
• Mip-Splatting Alias-free 3D Gaussian Splatting fixes some aliasing and dilation issues of original 3DGS paper, by effectively doing something smarter than “just add 0.3 here, lol” :)
• Compact3D: Compressing Gaussian Splat Radiance Field Models with Vector Quantization uses K-means clustering on colors, rotations, scales and spherical harmonics of 3DGS, to reduce data size.
• GaussianShader: 3D Gaussian Splatting with Shading Functions for Reflective Surfaces is another go at making gaussians relightable by estimating normal and other material parameters.
• Relightable Gaussian Codec Avatars is targeted at face avatars. They place gaussians on a coarse 3D mesh, and use a learnable radiance transfer with diffuse SH and specular SG. Clever way of plugging gaussian splats into an existing mesh-based avatar pipeline.
• SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering aligns gaussians closer with underlying surfaces, which then allows extracting the mesh, which then allows it to be animated or skinned. The final mesh still has gaussian splats “attached” to the polygons.

Unity Gaussian Splatting

The Unity Gaussian Splatting project that I created with intent of “eh, lemme try to make a quick toy 3DGS renderer in Unity, and maybe play around with data size reductions”, has somewhat surprisingly reached 1300+ GitHub stars. Since the previous blog post it got a bunch of random things:

• Support for HDRP and URP rendering pipelines in adition to the built-in one.
• Fine grained splat editing tools in form of selection and deletion (short video).
• High level splat editing tools in form of ellipsoid and box shaped “cutouts”. @hybridherbst did the initial implementation, and then shortly afterwards all other 3DGS editing tools got pretty much the same workflow. Nice!
  
• Ability to export modified/edited splats back into a .PLY file.
• Faster rendering via more tight oriented screenspace quads, instead of axis-aligned quads.
• I made the gaussian splat rendering+editing piece an actual package (OpenUPM page), and clarified license to be MIT.
• (not part of github release, but in latest main branch) More fine grained editing tools (move individual splats), ability to bake splat transform when exporting .PLY, and multiple splats can be merged together.

The project contains some bits that are not gaussian splat related, but might be useful elsewhere:

• A Unity port of AMD FidelityFX GPU radix sort (shader, C#).
• C# implementation of K-means clustering (source), with Burst and multi-threading.
• Editing tools implemented using Unity’s EditorToolContext/EditorTool framework (GaussianToolContext, GaussianTool, GaussianMoveTool).

Aaaand with that, I’m thinking that my toying around will end here. I’ve made a toy renderer and integration into Unity, learned a bunch of random things in the process, it’s time to call it a day and move onto something else. I suspect there will be another kphjillion gaussian splatting related papers coming out over the next year. Will be interesting to see where all of this ends up at!

Previous post was about making Gaussian Splatting data sizes smaller (both in-memory and on-disk). This one is still about the same topic! Now we look into clustering / VQ.

Teaser: this scene (garden tools from my own shed) is just 7.5 megabytes of data now. And it represents the metal shading (anisotropy / brushed metal parts) quite well!

Spherical Harmonics take up a lot of space!

In raw uncompressed Gaussian Splat data, majority of the data is Spherical Harmonics coefficients. If we ignore the very first SH coefficient (which we treat as “a color”), the rest is 45 floating point numbers for each splat (15 numbers for R,G,B channels each). For something like the “bike” scene with 6.1 million splats, this is 1.1GB data just for the SH coefficients alone. And while they can be converted into half-precision (FP16) floats with pretty much no quality loss at all, or into smaller quantized formats (Norm11 and Norm565 from previous post), that still leaves them at 350MB and 187MB worth of data. Even the idea that should not actually work – lay them out in a Morton order inside a texture and compress as GPU BC1 format – does not look entirely terrible, but is still about 46MB of data.

Are Spherical Harmonics even worth having? That’s a good question. Without them, the scenes still look quite good, but the surfaces lose quite a lot of “shininess”, especially different reflectance when moving the viewpoint. Below are “bike” and “garden” scenes, rendered with full SH data (left side) vs just color (right side):

How does “reflection” of the vase on the metal part of the table work, you might ask? The gaussians have “learned” the ages-old trick of duplicating and mirroring the geometry for reflection! Cute!

Anyway, for now let’s assume that we do want this “surface reflectivity looks nicer” effect that is provided by SH data.

Remember palettized images?

Remember how ages ago image files used to have a “color palette” of say 256 or 16 distinct colors, and each pixel would just say “yeah, that one”, pointing at the index of the color inside the palette. Heck, even whole computer displays were using palettes because “true color” was too costly at the time.

We can try doing the same thing for our SH data – given several million SH items inside a gaussian splat scene, can we actually pick “just some amount” of distinct SH values, and have each splat just point to the needed SH item?

Why, yes, we can. I’ve spent a bit of time learning about “vector quantization”, “clustering” and “k-means” and related jazz, and have played around with clustering SHs into various amounts (from 1024 up to 65536).

Note that SH data, at 45 numbers per splat, is quite “high dimensional”, and that has various challenges (see curse of dimensionality). One of them is that clustering millions of splats into thousands of items, in 45 dimensions, is not exactly fast. Another is that clustering might not produce good results. ⚠️ I don’t know anything about any of that; it could very well be that I should have done clustering entirely differently! But hey, whatever :)

Also, I’m very impatient, like if anything takes longer than 10 minutes I go “this is not supposed to be that long”. I first tried scikit-learn but that was taking ages to cluster SHs into even one thousand items. Faiss was way faster, taking about 5 minutes to cluster “bike” scene SHs into 16k items. However, I did not want to add that as a dependency, so I whipped up my own variant of mini-batch k-means using Burst’ed C# directly inside Unity. I probably did it all wrong and incorrectly, but it is about 3x faster than even Faiss and seems to provide better quality, at least for this task, so 🤷

So the process is:

• Take all the SH data from the gaussian splat scene,
• Cluster that into 4k - 16k distinct SH item “palette”. Store that. I’m storing as FP16 numbers, so that’s 360KB - 1.44MB data for the palette itself.
• For each original SH data point, find which item of the palette it is closest to. Store that inded per splat. I’m storing as 16 bits (even if some of the bits are not used), so for “bike” scene (6.1M splats) this is about 12MB indices.

Here’s full SH (left side) vs. SHs clustered into 16k items (right side):

This does retain the “shininess” effect, at expense of ~13MB data for either scene above. And while it does have some lighting artifacts, they are not terribly bad. So… probably okay?

Aside: the excellent gsplat.tech by Jakub Červený (@jakub_c5y) seems to also be using some sort of VQ/Clustering for the data. Seriously, check it out, it’s probably be nicest gaussian splatting thing right now w.r.t. usability – very intuitive camera controls, nicely presented file sizes, and works on WebGL2. Craftsmanship!

New quality levels

In my toy “gaussian splatting for Unity” implementation, currently I only do SH clustering at “Low” and “Very Low” quality levels.

Previously, “Low” preset had data sizes of 119MB, 49MB, 113MB; PSNR respectively 34.72, 31.81, 33.05):

Now, the “Low” preset clusters SH into 16k items. Data sizes 98MB, 41MB, 93MB; PSNR respectively 35.17, 35.32, 35.00:

The “Very Low” preset previously was pretty much unusable (data sizes of 74MB, 32MB, 74MB; PSNR 24.02, 22.28, 23.10):

However now the Very Low preset is in “somewhat usable” territory! File sizes are similar; the savings from clustered SH I’ve spent on other components that were suffering before. SH clustered into 4k items. Data sizes 79MB, 33MB, 75MB; PSNR 32.27, 30.19, 31.10:

Conclusions and future work

At this point, we can have “bike” and “garden” scenes in under 100MB of data (instead of original 1.4GB PLY file) at fairly acceptable quality. Not bad!

Of course gaussian splatting at this point is useful for “rotate around a scanned object” use case; it is not useful for “in games” or many other cases. We don’t know how to re-light them, or how to animate them well, etc. etc. Yet.

I haven’t done any of the “small things I could try” from the end of the previous post yet. So maybe that’s next? Or maybe look into how to further reduce the splat data on-disk, as opposed to just reducing the memory representation.

All the code for above is in this UnityGaussianSplatting#9 PR on github.

In the previous post I started to look at Gaussian Splatting. One of the issues with it, is that the data sets are not exactly small. The renders look nice:

But each of the “bike”, “truck”, “garden” data sets is respectively a 1.42GB, 0.59GB, 1.35GB PLY file. And they are loaded pretty much as-is into GPU memory as giant structured buffers, so at least that much VRAM is needed too (plus more for sorting, plus in the official viewer implementation the tiled splat rasterizer uses some-hundreds-of-MB).

I could tell you that I can make the data 19x smaller (78, 32, 74 MB respectively), but then it looks not that great. Still recognizable, but really not good (however, the artifacts are not your typical “polygonal mesh rendering at low LOD”, they are more like “JPG artifacts in space”):

However, in between these two extremes there are other configurations, that make the data 5x-10x smaller while looking quite okay.

So we are starting at 248 bytes for each splat, and we want to get that down. Note: everywhere here I will be exploring both storage and runtime memory usage, i.e. not “file compression”! Rather, I want to cut down on GPU memory consumption too. Getting runtime data smaller also makes the data on disk smaller as a side effect, but “storage size” is a whole another and partially independent topic. Maybe for some other day!

One obvious and easy thing to do with the splat data, is to notice that the “normal” (12 bytes) is completely unused. That does not save much though. Then you can of course try making all the numbers be Float16 instead of Float32, this is acceptably good but only makes the data 2x smaller.

You could also throw away all the spherical harmonics data and leave only the “base color” (i.e. SH0), and that would cut down 75% of the data size! This does change the lighting and removes some “reflections”, and is more visible in motion, but progressively dropping SH bands with lower quality levels (or progressively loading them in) is easy and sensible.

So of course, let’s look at what else we can do :)

Reorder and cut into chunks

The ordering of splats inside the data file does not matter; we are going to sort them by distance at rendering time anyway. In the PLY data file they are effectively random (each point here is one splat, and color is gradient based on the point index):

But we could reorder them based on “locality” (or any other criteria). For example, ordering them in a 3D Morton order, generally, makes nearby points in space be near each other inside the data array:

And then, I can group splats into chunks of N (N=256 was my choice), and hope that since they would generally be close together, maybe they have lower variance of their data, or at least their data can be somehow represented in fewer bits. If I visualize the chunk bounding boxes, they are generally small and scattered all over the scene:

This is pretty much slides 112-113 of “Learning from Failure” Dreams talk.

Future work: try Hilbert curve ordering instead of Morton. Also try “partially filled chunks” to break up large chunk bounds, that happen whenever the Morton curve flips to the other side.

By the way, Morton reordering can also make the rendering faster, since even after sorting by distance the nearby points are more likely to be nearby in the original data array. And of course, nice code to do Morton calculations without relying on BMI or similar CPU instructions can be found on Fabian’s blog, adapted here for 64 bit result case:

// Based on https://fgiesen.wordpress.com/2009/12/13/decoding-morton-codes/
// Insert two 0 bits after each of the 21 low bits of x
static ulong MortonPart1By2(ulong x)
{
    x &= 0x1fffff;
    x = (x ^ (x << 32)) & 0x1f00000000ffffUL;
    x = (x ^ (x << 16)) & 0x1f0000ff0000ffUL;
    x = (x ^ (x << 8)) & 0x100f00f00f00f00fUL;
    x = (x ^ (x << 4)) & 0x10c30c30c30c30c3UL;
    x = (x ^ (x << 2)) & 0x1249249249249249UL;
    return x;
}
// Encode three 21-bit integers into 3D Morton order
public static ulong MortonEncode3(uint3 v)
{
    return (MortonPart1By2(v.z) << 2) | (MortonPart1By2(v.y) << 1) | MortonPart1By2(v.x);
}

Make all data 0..1 relative to the chunk

Now that all the splats are cut into 256-splat size chunks, we can compute minimum and maximum data values of everything (positions, scales, colors, SHs etc.) for each chunk, and store that away. We don’t care about data size of that (yet?); just store them in full floats.

And now, adjust the splat data so that all the numbers are in 0..1 range between chunk minimum & maximum values. If that is kept in Float32 as it was before, then this does not really change precision in any noticeable way, just adds a bit of indirection inside the rendering shader (to figure out final splat data, you need to fetch chunk min & max, and interpolate between those based on splat values).

Oh, and for rotations, I’m encoding the quaternions in “smallest three” format (store smallest 3 components, plus index of which component was the largest).

And now that the data is all in 0..1 range, we can try representing it with smaller data types than full Float32!

But first, how does all that 0..1 data look like? The following is various data displayed as RGB colors, one pixel per splat, in row major order. With positions, you can clearly see that it changes within the 256 sized chunk (it’s two chunks per horizontal line):

Rotations do have some horizontal streaks but are way more random:

Scale has some horizontal patterns too, but we can also see that most of scales are towards smaller values:

Color (SH0) is this:

And opacity is often either almost transparent, or almost opaque:

There’s a lot of spherical harmonics bands and they tend to look like a similar mess, so here’s one of them:

Hey this data looks a lot like textures!

We’ve got 3 or 4 values per each “thing” (position, color, rotation, …) that are all in 0..1 range now. I know! Let’s put them into textures, one texel per splat. And then we can easily experiment with using various texture formats on them, and have the GPU texture sampling hardware do all the heavy lifting of turning the data into numbers.

We could even, I dunno, use something crazy like use compressed texture formats (e.g. BC1 or BC7) on these textures. Would that work well? Turns out, not immediately. Here’s turning all the data (position, rotation, scale, color/opacity, SH) into BC7 compressed texture. Data is just 122MB (12x smaller), but PSNR is a low 21.71 compared to full Float32 data:

However, we know that GPU texture compression formats are block based, e.g. on typical PC the BCn compression formats are all based on 4x4 texel blocks. But our texture data is laid out in 256x1 stripes of splat chunks, one after another. Let’s reorder them some more, i.e. lay out each chunk in a 16x16 texel square, again arranged in Morton order within it.

uint EncodeMorton2D_16x16(uint2 c)
{
    uint t = ((c.y & 0xF) << 8) | (c.x & 0xF); // ----EFGH----ABCD
    t = (t ^ (t << 2)) & 0x3333;               // --EF--GH--AB--CD
    t = (t ^ (t << 1)) & 0x5555;               // -E-F-G-H-A-B-C-D
    return (t | (t >> 7)) & 0xFF;              // --------EAFBGCHD
}
uint2 DecodeMorton2D_16x16(uint t)      // --------EAFBGCHD
{
    t = (t & 0xFF) | ((t & 0xFE) << 7); // -EAFBGCHEAFBGCHD
    t &= 0x5555;                        // -E-F-G-H-A-B-C-D
    t = (t ^ (t >> 1)) & 0x3333;        // --EF--GH--AB--CD
    t = (t ^ (t >> 2)) & 0x0f0f;        // ----EFGH----ABCD
    return uint2(t & 0xF, t >> 8);      // --------EFGHABCD
}

And if we rearrange all the texture data that way, then it looks like this now (position, rotation, scale, color, opacity, SH1):

And encoding all that into BC7 improves the quality quite a bit (PSNR 21.71→24.18):

So what texture formats should be used?

After playing around with a whole bunch of possible settings, here’s the quality setting levels I came up with. Formats indicated in the table below:

• F32x4: 4x Float32 (128 bits). Since GPUs typically do not have a three-channel Float32 texture format, I expand the data quite uselessly in this case, when only three components are needed.
• F16x4: 4x Float16 (64 bits). Similar expansion to 4 components as above.
• Norm10_2: unsigned normalized 10.10.10.2 (32 bits). GPUs do support this, and Unity almost supports it – it exposes the format enum member, but actually does not allow you to create texture with said format (lol!). So I emulate it by pretending the texture is in a single component Float32 format, and manually “unpack” in the shader.
• Norm11: unsigned normalized 11.10.11 (32 bits). GPUs do not have it, but since I’m emulating a similar format anyway (see above), then why not use more bits when we only need three components.
• Norm8x4: 4x unsigned normalized byte (32 bits).
• Norm565: unsigned normalized 5.6.5 (16 bits).
• BC7 and BC1: obvious, 8 and 4 bits respectively.

Here are the “reference” (“Very High”) images again (1.42GB, 0.59GB, 1.35GB data size):

The “Medium” preset looks pretty good! (280MB, 116MB, 267MB – 5.2x smaller; PSNR respectively 47.82, 48.73, 48.63):

At “Low” preset the color artifacts are more visible but not terribad (119MB, 49MB, 113MB – 12.2x smaller; PSNR respectively 34.72, 31.81, 33.05):

And the “Very Low” one mostly for reference; it kinda becomes useless at such low quality (74MB, 32MB, 74MB – 18.7x smaller; PSNR 24.02, 22.28, 23.1):

Oh, and I also recorded an awkwardly-moving-camera video, since people like moving pictures:

Conclusions and future work

The gaussian splatting data size (both on-disk and in-memory) can be fairly easily cut down 5x-12x, at fairly acceptable rendering quality level. Say, for that “garden” scene 1.35GB data file is “eek, sounds a bit excessive”, but at 110-260MB it’s becoming more interesting. Definitely not small yet, but way more within being usable.

I think the idea of arranging the splat data “somehow”, and then compressing them not by just individually encoding each spat into smaller amount of bits, but also “within neighbors” (like using BC7 or BC1), is interesting. Spherical Harmonics data in particular looks quite ok even with BC1 compression (it helps that unlike “obviously wrong” rotation or scale, it’s much harder to tell when your spherical harmonics coefficient is wrong :)).

There’s a bunch of small things I could try:

• Splat reordering: reorder splats not only based on position, but also based on “something else”. Try Hilbert curve instead of Morton. Try using not-fully-256 size chunks whenever the curve flips to the other side.
• Color/Opacity encoding: maybe it’s worth putting that into two separate textures, instead of trying to get BC7 to compress them both.
• I do wonder how would reducing the texture resolution work, maybe for some components (spherical harmonics? color if opacity is separate?) you could use lower resolution texture, i.e. below 1 texel per splat.

And then of course there are larger questions, in a sense of whether this way looking at reducing data size is sensible at all. Maybe something along the lines of “Random-Access Neural Compression of Material Textures” (Vaidyanathan, Salvi, Wronski 2023) would work? If only I knew anything about this “neural/ML” thing :)

All my code for the above is in this PR on github (merged 2023 Sep).

In the followup post I look at making them even smaller!

SIGGRAPH 2023 just had a paper “3D Gaussian Splatting for Real-Time Radiance Field Rendering” by Kerbl, Kopanas, Leimkühler, Drettakis, and it looks pretty cool! Check out their website, source code repository, data sets and so on (I should note that it is really, really good to see full source and full data sets being released. Way to go!).

I’ve decided to try to implement the realtime visualization part (i.e. the one that takes already-produced gaussian splat “model” file) in Unity. As well as maybe play around with looking at whether the data sizes could be made smaller (maybe use some of the learnings from float compression series too?).

What’s a few million badly rendered boxes among friends, anyway?

For the impatient: I got something working over at aras-p/UnityGaussianSplatting, and will tinker with things there some more. And since this post, I wrote several others:

• Making Gaussian Splats smaller,
• Making Gaussian Splats more smaller
• Gaussian Explosion

Meanwhile, some quick random thoughts on this Gaussian Splatting thing.

What are these Gaussian Splats?

Watch the paper video and read the paper, they are pretty good!

I have seen quite many 3rd party explanations of the concept at this point, and some of them, uhh, get a thing or two wrong about it :)

• This is not a NeRF (Neural Radiance Field)! There is absolutely nothing “neural” about it.
• It is not somehow “fast, because it uses GPU rasterization hardware”. The official implementation does not use the rasterization pipeline at all; it is 100% done with CUDA. In fact, it is fast because it does not use the fixed function rasterization, as we’ll see below.

Anyway,

Gaussian Splats are, basically, “a bunch of blobs in space”. Instead of representing a 3D scene as polygonal meshes, or voxels, or distance fields, it represents it as (millions of) particles:

• Each particle (“a 3D Gaussian”) has position, rotation and a non-uniform scale in 3D space.
• Each particle also has an opacity, as well as color (actually, not a single color, but rather 3rd order Spherical Harmonics coefficients - meaning “the color” can change depending on the view direction).
• For rendering, the particles are rendered (“splatted”) as 2D Gaussians in screen space, i.e. they are not rendered as scaled elongated spheres, actually! More on this below.

And that’s it. The “Gaussian Splatting” scene representation is just that, a whole bunch of scaled and colored blobs in space. The genius part of the paper is several things:

• They found a way how to create these millions of blobs that would represent a scene depicted by a bunch of photos. This is using gradient descent and “differentiable rendering” and all the other things that are way over my head. This feels like the major contribution, like maybe previously people assumed that in order for gradient descent optimizer to work nicely, you need to use a continuous or connected scene representation (vs “just a bunch of blobs”), and this paper proved that wrong? Anyway, I don’t understand this area, so I won’t talk about it more :)
• They have developed a fast way to render all these millions of scaled particles. This by itself is not particularly ground breaking IMHO, various people have noticed that using something like a tile-based “software, but on GPU” rasterizer is a good way to do this.
• They have combined existing established approaches (like gaussian splatting, and spherical harmonics) in a nice way.
• And finally, they have resisted the temptation to do “neural” anything ;)

Previous Building Blocks

The Gaussian Splatting seems to be invented around year 2001-2002, see for example “EWA Splatting” paper by Zwicker, Pfister, Van Baar, Gross. There they have scaled and oriented “blobs” in space, calculate how would they project onto screen, and then do the actual “blob shape” (a “Gaussian”) in 2D, in screen-space. A bunch of signal processing, sampling, aliasing etc. math presumably supports doing it that way.

Speaking of ellipsoids, Ecstatica game from 1994 had a fairly unique ellipsoid-based renderer.

Spherical Harmonics (a way to represent a function over a surface of a sphere) have been around for several hundred years in physics, but really were popularized in computer graphics around 2000 by Ravi Ramamoorthi and Peter-Pike Sloan. But actually, a 1984 “Ray tracing volume densities” paper by Kajiya & Von Herzen might be the first use of them in graphics. A nice summary of various things related to SH is at Patapom’s page.

Point-Based Rendering in various forms has been around for a long time, e.g. particle systems were used since “forever” (but typically used for vfx / non-solid phenomena). “The Use of Points as a Display Primitive” is from 1985. “Surfels” paper is from 2000.

Something closer to my heart, a whole bunch of demoscene demos are using non-traditional rendering approaches. Fairlight & CNCD have several notable examples, e.g. Agenda Circling Forth (2010), Ceasefire (all falls down..) (2010), Number One / Another One (2018):

Real-time VFX tools like Notch have pretty extensive features for creating, simulating and displaying point/blob based “things”.

Ideas of representing images or scenes with a bunch of “primitive shapes”, as well as tools to generate those, have been around too. E.g. fogleman/primitive (2016) is nice.

Media Molecule “Dreams” has a splat-based renderer (I think the shipped version is not purely splat-based but a combination of several techniques). Check out the most excellent “Learning from Failure” talk by Alex Evans: at SIGGRAPH 2015 (splats start at slide 109) or video from Umbra Ignite 2015 (splats start at 22:34).

Tiled Rasterization for particles has been around at least since 2014 (“Holy smoke! Faster Particle Rendering using Direct Compute” by Gareth Thomas). And the idea that dividing screen into tiles, doing a bunch of things “inside the tile” thus cutting on memory traffic, is how entire mobile GPU space operates, and has been operating since “forever”, tracing back to first PowerVR designs (1996) and even Pixel Planes 5 from 1989.

This is all great! Taking existing, developed, solid building blocks and combining them in a novel way is excellent work.

My Toy Playground

My current implementation (of just the visualizer of Gaussian Splat models) for Unity is over at github: aras-p/UnityGaussianSplatting. Current state is “it kinda works, but it is not fast”:

• The rendering does not look horrible, but does not exactly match official implementation. Here is official vs my rendering of the same scene. Official one has more small detail, and lighting is slightly different Fixed!
  
• Performance is not great. The scene above renders on NVIDIA RTX 3080 Ti at 1200x800 in 7.40ms (135FPS) in the official viewer, whereas my attempt is 23.8ms (42FPS) currently, i.e. 4x slower. For sorting I’m using some fairly simple GPU bitonic sort (official impl uses CUDA radix sort which is based on OneSweep algorithm). Rasterization in their case is tile-based and written in CUDA, whereas I’m “just” using regular GPU rasterization pipeline and rendering each splat as a screenspace quad.
  • On the plus side, my code is all regular HLSL within Unity, which means it also happens to work on e.g. Mac just fine. The scene above on Apple M1 Max renders in 108ms (9FPS) though :/
  • My implementation seems to use 2x less GPU memory right now too (official viewer: 4.8GB, mine: 2.2GB and that’s including whatever Unity editor takes).

So all of that could be improved and optimized quite a bit!

One thing I haven’t seen talked much about, by everyone super excited about Gaussian Splats, is data size and memory usage. Yeah, rendering is nice, but this bicycle scene above is 1.5GB of data on-disk, and then at runtime it needs some more (for sorting, tile based rendering etc.). That scene is six million blobs in space, with each of them taking about 250 bytes. There has to be some way to make that smaller! Actually the Dreams talk above has some neat ideas.

Maybe I should play around with that. Someday!