Recently I’ve been playing with 3D Gaussian Splatting.
First, I take 337 images of our son’s stroller from different angles.
Then, I use colmap to infer camera poses and construst sparse point cloud from the images.
Latly, I use the 3DGS official repo to generate the stroller’s 3D model. It takes about a couple dozen minutes for 3DGS to finish training (173 million splats are generated), and the result pretty amazing!