A Driving World Model based on SVD (incl. all training code)

[Little-Podi]

We are happy to share our recent work, which is an advanced world model for autonomous driving. Developed upon SVD, our model possesses long-term prediction ability and multi-modal action controllability: https://arxiv.org/abs/2405.17398

For those who are also enthusiastic about SVD, we have implemented training and sampling code. Hope these are helpful: https://github.com/OpenDriveLab/Vista

We really appreciate the generous efforts of Stability AI. We cannot realize our ideas without this excellent codebase! :heart: