Datasets:

Languages:
English
ArXiv:
License:

Update Metadata for Assets

#18
by krishmehta0 - opened

Hi,

This dataset is extremely valuable, but hard to use because of the lack of metadata, along with annotation categories for LVIS being too broad. I'd love to see how I could contribute to setting up a scalable pipeline that uses VLMs to create better metadata and descriptions for object files, and potentially also allows to update the thumbnail images in the process. This would significantly improve the usability of this collection, and allow for better usage across different industries.

(Agree this is a very powerful dataset)

Sign up or log in to comment