The fastest-moving corner of the field. Neural reconstruction and generative models now turn a single image — or a text prompt — into geometry, while classical photogrammetry remains the accuracy benchmark.
Single-image to 3D mesh in under a second on a GPU — the model that made image-to-3D practical.
Free photogrammetry on the AliceVision pipeline — turn a photo set into a dense, textured model.
Tencent's open-weights generator — high-fidelity textured meshes from image or text.
The structure-from-motion reference — the accuracy backbone behind much of NeRF and gaussian-splat work.