17h ago

Waymo and XPENG use generative video models to synthesize multi-camera training data from a single front-facing dashcam

2 top authors

This approach programmatically scales training datasets using online footage

Original post

An emerging application of video generative models for autonomous driving to augment data: given a dashcam/frontcam video, generate the surround camera views video. Huge amounts of such videos online or at specific suppliers. Both Waymo and Xpeng showed such approaches #cvpr2026

2:57 PM · Jun 3, 2026

Sentiment

Pos100%

Neg0%

Users appreciate the Waymo and XPENG showcase of generative models creating surround views from dashcam video because it offers a clear, easy-to-use example of synthetic data for explanations.

1 comment with sentiment.

1 more post

Lucas Beyer (bl16)@giffmana·8hReply
@abursuc This is a very nice and easy to convey example of synthetic data, thanks for sharing I'll definitely use it in some future explanations !
View on