3 ms·
I didn’t expect Gaussian splats to be so good at approximating geometry. It’s cool when you see a new foundational approach to something that’s been done a cert
by randall 3y ago
I didn’t expect Gaussian splats to be so good at approximating geometry. It’s cool when you see a new foundational approach to something that’s been done a certain way for decades.
- eurekin 3y agoSomeone mentioned previously that guassian splats are ideal extension of Ai image generation into 3d and they might have turned out correct
- kridsdale1 3y agoThey’re basically voxels after we abandon Cartesian 3-dimensionality (which imposes cubes)
- porphyra 3y agoMore like point clouds since they are not constrained to a grid. NERF is more like voxels though.
- blovescoffee 3y agoWhat do you mean NERF is like voxels? It is very much NOT like voxels
- porphyra 3y agoThe original NERF paper had a voxel-like grid, but instead of regular voxels with a single color per grid cell, they had an MLP in each cell that models the color as a function of position and angle.
- blovescoffee 3y agoThe literal definition of a NeRF and rendering function in the original paper are a continuous non-discretized vector-field and sampling based render function.
- porphyra 3y agoI'm sorry, you're right. I got confused with some other papers like Plenoxels that use voxels.
- blovescoffee 3y agoMost of the time they aren't that good at approximating geometry. They are good at approximating the "appearance" of geometry. However, many regularization techniques and priors can be introduced to make the Gaussian splatting technique better at geometry approximation.