Battlefields are full of uncertainty and mistakes. Omnivision provides clarity. This project builds a 3D model from a peer-to-peer network of cameras working together. Clients can see the output of any other camera on the network. Multiple video feeds can be put together to generate a single, evolving 3D model.

Omnivision showcase

How it's made

3D Model: Used COLMAP to align videos, infer positional data, and build structure-from-motion (SFM) models. Used OpenMVS to create a point cloud and generate the 3D model. Camera Network: Video is created and stored in chunks. IPFS is used to host and distribute the video chunks. libp2p pubsub is used to propagate content addresses.