This question was the topic of a recent town hall hosted by the Metaverse Standards Forum.
Throughout the geospatial, AEC, and 3D visualization space, perhaps the biggest topic of conversation over the last 12 months has been Gaussian Splats. Still an extremely new area, even within the context of this 3D visualization landscape that is constantly changing, this rendering technique for 3D scenes is capturing the attention and imagination of professionals across numerous industries, from entertainment to AEC.
The first paper introducing Gaussian Splatting was released less than two years ago, but already we’re seeing tons of early adoption among innovators and companies starting to experiment with how they can incorporate the technique into their offerings. XGRIDS, for example, is fusing splats with their handheld laser scanners. We will almost certainly see more companies join that trend over the next year.
This rapid adoption and overall excitement around the technique is invigorating in a lot of ways, and people are running to create interesting use cases and products based on the original paper. Last year, Geo Week News held a webinar on this topic that ran through some of these products and use cases, for example. However, on the flip side of this invigoration comes questions about how to best harness the abilities of the technique. Specifically, those across the industry are starting to wonder if now is the time to start thinking about standardization around Gaussian Splatting.
Earlier this year, many of the innovators on the cutting edge in this area came together for a big town hall put together by the Metaverse Standards Forum to explore this very question. Patrick Cozzi, the founder of Cesium and Chief Platform Officer with Bentley Systems, and Alexey Medvedev, AR Tech Lead at Meta, hosted the event, which featured a number of speakers across many different use cases and industries, with sections of the conversation including foundations, momentum, web, and geospatial & digital twins.
The entire three-and-a-half hour conversation has been posted on YouTube for free viewing, and it’s worth watching the entire discussion. Presenters talked a lot about the evolution of Gaussian Splatting, explained exactly how it works, and showed off some great use cases that are already in place. Additionally, the group is putting on a part two of the discussion on Wednesday from 12:30 PM – 2:00 PM Eastern Time.
As we noted, the conversation is wide-ranging, but it’s worth focusing specifically on the question in the title of the town hall: Is the Gaussian Splatting space ready for standardization? There wasn’t a real conclusion to this question in the conversation, which shouldn’t come as a surprise given how new the space is. The benefits of standardization are clear, though.
Much of the conversation was around data types and sharing, which is important for this topic; one of the biggest benefits of Gaussian Splats right now is creating easily shareable visualizations. That, of course, lends itself to collaborative workflows, but if people are working with different file formats without a standardized path, that collaboration becomes much more difficult, if not impossible. There’s no doubt or even disagreement as to whether or not standardization needs to come at some point – it certainly does.
The question is whether or not the industry is ready for it today, and there were plenty of compelling arguments against that idea, even among those who believe standardization needs to happen eventually. To boil it down, the crux of the argument against standardization right now is that the space is moving too quickly. Multiple speakers noted that there are new developments coming out every day, and while that may be an exaggeration, it’s not by as much as you would think. There is real concern that creating a standard right now, while the entire ecosystem is very much in flux, could stifle innovation and paint the space into an unnecessary corner.
Ultimately, there was no real conclusion to this conversation, which makes sense given the aforementioned rapid development in this space. That being said, it does seem like there was broad agreement that now is at least the time to start thinking about these standardizations. Creating a standard is not something that happens overnight, and given how many industries Gaussian Splatting has the potential to touch, it will take a long time to get every sector on board for a standard. We may not be in the place right now where a standard makes sense, but we’re likely to reach that point within the next few years, and it’s good to be prepared now for that future.
