Anything-3D

Why we included this project

For anyone tinkering with 3D reconstruction, this repo shows how a modern segmentation model like Segment Anything can plug into established 3D pipelines. It bundles demos that turn a single image into a segmented 3D object, generate novel views, and build NeRF reconstructions, plus an experimental face-reconstruction path. The code is split into small, self-contained demo folders, so it reads like a tour of possible architectures rather than one polished product. That makes it a good starting point for researchers, grad students, or hobbyists who want to compare approaches and experiment with combining 2D segmentation and 3D lifting. Expect to adapt code to your own data and dependencies, since this is an academic demo, not a maintained service.