InVesta VisionbySalih İbrahimbaş·Oct 12The Missing Views Problem: Can AI Reconstruct It Has Never Seen?On sparse-view reconstruction, generative 3D, and the uncomfortable difference between a convincing scene and a correct one.
Papers in 100 Lines of Code·Sep 21AnySplat Turns 32 Unposed Photos Into a 3D Scene in 1.4 SecondsBuilt on VGGT and trained without 3D ground truth, it predicts cameras, depth, and millions of 3D Gaussians in one forward pass — and…A response icon1A response icon1
Papers in 100 Lines of Code·Sep 14MVSplat Beat pixelSplat With 10× Fewer Parameters. Here’s How.Same benchmark, opposite ideas: pixelSplat learned to predict depth, while MVSplat measured it by matching views — and won on quality…A response icon1A response icon1
Papers in 100 Lines of Code·Sep 11Apple Turns One Photo Into 1.2 Million 3D GaussiansSHARP reconstructs a metric-scale 3D scene in under a second, using a single forward pass instead of diffusion or per-scene optimization.A response icon1A response icon1
Papers in 100 Lines of Code·Sep 8Do World Models Replace 3D Reconstruction?Genie 3, Marble, and SHARP can all turn images into explorable worlds — but they disagree on something fundamental: whether a scene should…
Papers in 100 Lines of Code·Sep 4Depth Anything 3 Just Made VGGT Look OverengineeredA plain DINO transformer with one depth-and-ray prediction target beats VGGT on camera pose and geometry — and can even produce 3D…A response icon1A response icon1
Papers in 100 Lines of Code·Sep 1One Transformer Just Replaced the Classical 3D Vision StackVGGT predicts camera poses, depth maps, and dense 3D point clouds from hundreds of images in a single forward pass — no…A response icon2A response icon2
Papers in 100 Lines of Code·Aug 273D Gaussian Splatting No Longer Has to Start From ScratchFeed-Forward 3D Gaussian Splatting: From Proof of Concept to ProductionA response icon1A response icon1
Pann·Aug 7From 2D Images to 3D Reconstruction: My First Experiment with VGGTRecently, I experimented with VGGT (Visual Geometry Grounded Transformer) to see how well it can reconstruct a 3D scene from 2D images. For…
