Across the long practice of turning photographs into three-dimensional maps, a newer method called Gaussian splatting has begun to draw attention as a possible companion to, or rival of, the textured meshes that photogrammetry has produced for years. A recent webinar hosted by SimActive, a mapping software maker, examines where this approach fits in real work, from how images must be captured to whether the results serve visualization or measurement. The material describing that session does not announce a verdict, and it comes from a party with a stake in the answer, so the question of which
Gaussian Splatting vs. Textured Meshes: Where the New 3D Approach Fits Real-World Workflows
The piece poses the question and leaves the answer to the recording.
Mark
So what actually happened here? A webinar?
Mimi
Yes. Dr. Jules Morel of Mirukuru in Japan met with SimActive's team to discuss how Gaussian splats are made and how they compare with textured meshes. This piece is the announcement of that recording.
Mark
What does the webinar cover?
Mimi
Four things: the fundamentals, capturing and processing imagery for good results, using splats for visualization and measurement, and comparing them with meshes.
Mark
Why does the comparison matter?
Mimi
Meshes are the established photogrammetry output. Splats are newer, so people want to know where each belongs in practice.
Luke
Does the piece say splats beat meshes?
Mimi
No. It poses the question and leaves the answer to the recording.
Luke
Then I'd flag that nothing here is a finding. It's a description of a session. No accuracy numbers, no tests.
Mark
And SimActive's role?
Mimi
They make Correlator3D, which already produces splats and meshes alike.
Luke
Which means they have a commercial interest in both outputs looking useful. That's not damning, but it's worth knowing.
Mark
So what should a reader take away?
Mimi
That splats are being weighed by application, and that the real evidence lies in the webinar and in independent testing.
O Pulso
- A new way of rendering realistic 3D scenes from ordinary imagery is challenging the textured meshes that mapping professionals have relied on for years.
- The question on the table is whether splats will replace meshes outright or find their place beside them, and the answer is being left to specific applications.
- Practitioners are turning to a webinar to learn what capture and processing demands high-quality splats require, and whether they can support measurement as well as visualization.
- The source is a software vendor that already sells splats alongside meshes and offers no independent benchmarks, so claims about accuracy remain untested in the material itself.
Across the long practice of turning photographs into three-dimensional maps, a newer method called Gaussian splatting has begun to draw attention as a possible companion to, or rival of, the textured meshes that photogrammetry has produced for years. A recent webinar hosted by SimActive, a mapping software maker, examines where this approach fits in real work, from how images must be captured to whether the results serve visualization or measurement. The material describing that session does not announce a verdict, and it comes from a party with a stake in the answer, so the question of which representation will endure remains open, settled less by claims than by how each tool performs for the particular task at hand.
Dr. Jules Morel, a data analyst at Mirukuru in Japan, recently joined the team at SimActive for a webinar on a question now circulating through 3D mapping: can Gaussian splatting, a newer way of building realistic scenes from imagery, stand alongside the textured meshes photogrammetry has long produced?
Spatial Source published a short write-up pointing readers to the recording. It describes what the session covers rather than what it concluded. The webinar moves from the fundamentals of splatting to capture and processing requirements, then to visualization and measurement uses, and finally to a comparison of splats and textured meshes by application. The framing invites the question without answering it.
SimActive is an interested party. Its Correlator3D software already generates point clouds, terrain models and 3D models from aerial and satellite imagery, and now includes Gaussian splats among its outputs. The company sells to mapping firms and government organizations, and the write-up includes no independent benchmarks, accuracy figures or side-by-side tests.
What the material does establish is that the industry increasingly judges splats and meshes by use rather than by novelty. Readers who want evidence should watch the recording and look for independent testing of splats in measurement work.
Dr. Jules Morel, a Data Analyst at Mirukuru in Japan, recently sat down with the team at SimActive to talk about a question that is making the rounds in 3D mapping: can Gaussian splatting, a newer way of building realistic 3D scenes from imagery, stand alongside the textured meshes that photogrammetry has long produced?
The conversation was recorded as a webinar, and Spatial Source has published a short write-up pointing readers to it. The write-up is a description of the webinar, not a report on its findings, so what follows is what it says the session covers rather than what the session concluded.
According to the description, the webinar has four parts. It begins with the fundamentals of Gaussian splatting. It then turns to how imagery should be captured and processed to get high-quality results. It goes on to look at splats for two purposes, visualization and measurement. Finally it compares splats with traditional textured meshes and the applications each suits.
The framing matters. The description presents splatting as an approach gaining attention for producing realistic 3D representations from images, and it asks where the method fits in real-world workflows. It does not announce a winner. The title's question, whether splatting will replace textured meshes, is left for the webinar itself, and the summary offers no answer.
SimActive is a party with an interest here. The company develops Correlator3D, a patented photogrammetry package that generates geospatial data from satellite and aerial imagery, including drone captures. By its own description, the software performs aerial triangulation and produces dense digital surface models, digital terrain models, point clouds, orthomosaics, 3D models and vectorized 3D features. Gaussian splats are now on that list too. The company says the software uses GPU technology and multi-core CPUs to process large datasets quickly, and that it sells to mapping firms and government organizations worldwide.
That context shapes how to read the piece. It comes from a software vendor that already offers splats alongside meshes, and it includes no independent benchmarks, accuracy figures or side-by-side test results. Claims about quality or measurement reliability are therefore not demonstrated in the text.
What the material does establish is that the industry is treating splats and meshes as outputs to be compared by use, not simply as old versus new. Whether splats replace meshes or sit beside them will depend on specific applications. Readers who want evidence should watch the webinar recording itself and look for independent testing of splats for measurement work.
Citações Notáveis
Gaussian splatting is gaining attention as a new approach to creating realistic 3D representations from imagery, and the open question is where it fits in real-world workflows.— Paraphrase of the SimActive webinar description
Quer a matéria completa?
Leia o original em Spatial Source ↗
General News