Apple’s ML team has been surprisingly consistent with these "quiet" open-source releases lately (OpenELM, MGIE, and now DSplats). Using 3D Gaussian Splatting (3DGS) as the representation layer for a diffusion model feels like the right move over the older NeRF-based approaches.
Discussion (503 points, 2 days ago, 107 comments) https://news.ycombinator.com/item?id=46284658
Apple’s ML team has been surprisingly consistent with these "quiet" open-source releases lately (OpenELM, MGIE, and now DSplats). Using 3D Gaussian Splatting (3DGS) as the representation layer for a diffusion model feels like the right move over the older NeRF-based approaches.