Hi Cameroncone - this is seriously impressive work. I’d love to see an in-depth video breakdown of the entire workflow and each stage of the process. Even a simplified version using just 1–2 rooms would be amazing. I think loads of us would genuinely love to watch that kind of content and learn from the pipeline you’ve built.
Yesss. Please do a tutorial. I'll check this every day for years if I have to until you do. You've nearly mastered many parts of this use of Gaussian Splatting. At first, I was like: maybe the model is too small- I feel unnaturally tall walking around here. That was about 3.7 seconds before I remembered I'm 5'5"
Amazing work! How did you solve a problem of white walls? With my scans reality capture or postshot cannot establish positions of walls due to lack of features. On your splat it looks amazing tho.
Hey, I’m trying to replicate this as a little project because I really love it, but I’m struggling with the initial capture/shooting part.
I’ve tried using apps like Scaniverse and Polycam Pro, then exporting PLY files to import into SuperSplat Editor afterwards, but it doesn’t seem to accept the files — it says there’s no Gaussian splatting data.
Would really appreciate any help.
Here are some of the tools I used: 5D Mark 4, 14mm and 35mm lenses, DJI Pocket Cam, Reality Capture and Post Shot, Super Splat, and a little bit of C4D for object cleanup.
I separately captured each area, rebuilt them in Super Splat to compress the files, and finally combined them all into one massive splat. This pipeline has been my endeavor for the past few months. I have tried to create a reliable pipeline by developing multiple versions of this house.
The walk mode collision data piqued my interest the most.
The long-term objective is to discover the optimal combination of quality and timing to successfully implement this feature in real estate.