Find a Dataset on Universe
Learn how to find data to use in your vision projects on Universe.
Last updated
Was this helpful?
Learn how to find data to use in your vision projects on Universe.
Last updated
Was this helpful?
To find a dataset on Universe, type in what you want to identify into the search bar.
A list of results will appear with datasets semantically related to your query:
By default, Universe uses a CLIP-based semantic search to find datasets that match your query. If you want to search by keyword, toggle the "Search By" option and choose "Metadata".
You can filter search results by:
Datasets that have a model
Project type (object detection, classification, segmentation, keypoint detection, multimodal.)
The type of model trained (i.e. YOLOv12)
There are also advanced filters that you can use to further refine your search by the number of images in a dataset, the number of stars the project has, and more.
You can click on an individual dataset to find out more about the data.
To find out how a dataset is licensed, scroll to the "Cite this Project" section.
The license will appear next to the "License" label. If no license is specified, this means the author reserves all rights.
The contents of the Cite This Project tab will contain the information you need to cite the project.
Once you have found a dataset you want to use, you can fork it into your account.
The Classes list tells you what classes have been used as labels in the dataset. if you import the dataset into your workspace.