Datasets

Organize your data and labels with datasets

Datasets are a way to organize your assets and labels. After labeling they can be exported in a variety of formats to be used in Machine Learning or shared with others.

If you are a HyperLabel user and are moving projects from HyperLabel, you may want to check out our HyperLabel to Sense transition guide.

Since labeled data is exported by dataset, we recommend making different datasets for different machine learning objectives. For instance, you may have one dataset for detecting boats in images, and another for reading chess boards.

Viewing Datasets

To go view and create datasets, click Data Labeler in the side nav under Label & Train

  1. To open an existing dataset click on its preview card.

  2. To navigate back to the dataset list from any page click Data Labeler tab.

Create a Dataset

  1. To create a new dataset, click on the “Create Dataset” button near the top right corner.

  2. Enter Name the project and click "Save & Continue to Sources"

  3. You will be taken to the Sources tab. From here you can add your data sources

Import Dataset

Importing is used to load a dataset that has been exported in the HyperLabel/Sense JSON format.

  1. Click "Import Dataset"

  2. Navigate to where the project JSON file is saved

  3. Select JSON file and click "open"

When the import is complete, a notification badge will appear and the dataset will appear on Data Labeler page with a preview card.

As of this time, Sense only supports importing datasets in the HyperLabel and Sense JSON formats