Less Is More : Scalable Visual Navigation from Limited Data

[qinhaijian23]

Hi @JonasFrey96 and the team,

First of all, congratulations on your recent paper
"Less Is More: Scalable Visual Navigation from Limited Data"
! I really enjoyed reading it.

I am very interested in reproducing your results and trying out the LIMO policy. The paper mentions that the code and models will be released. Since this work is built upon the GrandTour dataset, I was wondering if the LIMO code will be hosted here, or if there will be a separate repository for it?

If it's not released yet, is there a rough timeline for when we might expect it?

Thanks a lot for your amazing work!

Best regards

[yv1es]

Dear @Qhijane,

Thank you for your kind message. We’re glad you enjoyed the paper!

In the coming days, we’ll be launching the project website, and I’ll share the link here once it’s live. We’ll also add the datasets to the GrandTour Hugging Face dataset repository next week.

The LiMo code and checkpoints will be released in a separate repository. It will be linked on the project website, and I’ll share the link here as well. I’m currently cleaning up the codebase; you can expect it to be public within the next one to two weeks.

Best regards

[qinhaijian23]

Dear @Qhijane,

Thank you for your kind message. We’re glad you enjoyed the paper!

In the coming days, we’ll be launching the project website, and I’ll share the link here once it’s live. We’ll also add the datasets to the GrandTour Hugging Face dataset repository next week.

The LiMo code and checkpoints will be released in a separate repository. It will be linked on the project website, and I’ll share the link here as well. I’m currently cleaning up the codebase; you can expect it to be public within the next one to two weeks.

Best regards

Thank you for the detailed update! That sounds great. I'm really looking forward to the release and will keep an eye on this thread for the links. Good luck with the codebase cleanup!

[yv1es]

Dear @qinhaijian23 ,

Thanks for your patience.

You can find the project page here: https://leggedrobotics.github.io/less-is-more
And the code repository here: https://github.com/leggedrobotics/less-is-more

The MPPI planner used to build the dataset will follow, but you can now find everything to load our dataset, to train the model and to run inference there.

Best regards