3,072 karma · joined June 28, 2013
---
X-Maps: Using a tiny laser projector and an event camera to estimate depth at 60 Hz on a laptop CPU. The algorithm is using only the time stamps of when the laser passes over the scene, not its color or intensity, so you can project any content you like. Fun for AR demos!
https://fraunhoferhhi.github.io/X-maps/
---
Self-Organizing Gaussian Grids: 3D data is awkward to compress. But there's plenty of solutions for compressing 2D data (images!). So let's organize our 3D data into a 2D grid, where grid neighbors are also close in 3D. That's a hard problem, but we can leverage novel assignment algorithms with GPU power for parallelizing that, to get the sorting done in a few seconds.
https://fraunhoferhhi.github.io/Self-Organizing-Gaussians/
---
https://www.3dgs.zip/ resources on 3D Gaussian Splatting from our research group.
https://survey.3dgs.zip/ - a comparison of different compression methods for 3D Gaussian Splatting scenes.