Code & tutorial release

[Jueqian]

Thanks for your great work! Has the complete code and tutorial been released now?

[Czm369]

The tutorial, pretrained weight for stage 1&2 and inference code will be released before Oct 1st, 2025.

[scott198510]

@Czm369
Does the "multi-view" here include fisheye cameras, or is it limited to only pinhole cameras?
As is well known, the image generation based on fisheye camera setups differs significantly from that of pinhole cameras.

[Czm369]

Our method constructs BEV features based on the spatial mapping defined by the camera intrinsics and extrinsics. In theory, it can support input from fisheye cameras, as long as the spatial relationships are correctly established using the fisheye camera parameters.

[Orbis36]

Hello there, thanks for your work!
Oct 1 is close and looking forward to seeing your codes!

[LongType81818]

is there any update on the release date?

[Czm369]

As the paper was rejected by ICCV, NeurIPS, and ICLR, the open-source plan was temporarily suspended. To further promote progress in the field, we will release the inference code within one week and provide an updated usage tutorial.

[Czm369]

[csj777]

啊这，之前看意见以为稳中iclr了

[Czm369]

[Czm369]

We have released the inference code and pretrained weights for BEV-VAE. The tutorial is coming soon! 🚀🚀🚀

[facethewindha]

Hello, may I ask if you plan to open-source the training code in the future?

[Czm369]

Hello, may I ask if you plan to open-source the training code in the future?

The training code will be made open-source later, but it is only available for academic research purposes. This is because I am currently working on the extension and commercialization of BEVVAE.