Web Demo v2
KoroKoro is an automated pipeline for converting 2D videos into detailed 3D models using advanced techniques.
Scroll Down
Scroll Down
Scroll Down
Scroll Down
Scroll Down
Scroll Down
Scroll Down
Video Ingestion & Processing
Given an input video, 40 frames are extracted, these 40 frames are processed with NerfStudio which uses Colmap under the hood to generate the poses.
Image Transformation
Grounding DINO is used to create a bounding box around an object of interest in the scene, this bounding box is then passed to SegmentAnything2 to generate a mask around the object. The mask is then used to crop the object from the image.
3D Reconstruction
We train a Splatfacto given the processed inputs from the previous steps to generate a 3D reconstruction of the object.
Ready!
Input
KoroKoro Version One
KoroKoro Version Two