Branch
main branch https://mmdetection3d.readthedocs.io/en/latest/
📚 The doc issue
Hi, thank you for releasing 'EmbodiedScan v2 beta'. But I have a question.
In EmbodiedScan v1 dataset files, it has:

But In EmbodiedScan v2 dataset files, it has:

So, how could I train a 3D grounding task with a mini-dataset in the v2 dataset?
Suggest a potential alternative/fix
No response