Data Explorer¶
The Data Explorer is a powerful tool within the ML cube Platform that allows you to explore, visualize, and analyze your data batches. It provides a flexible interface to inspect your data samples, understand their properties, and prepare for model retraining or analysis.
The Data Explorer is designed to work with various data types, including tabular data, images and text.
You can navigate to the Data Explorer from your task's menu.
Data Views¶
The Data Explorer offers two main ways to visualize your data:
- Sample View: This view presents your data in a paginated table, showing key information for each sample, such as its metadata and any associated predictions or drift scores. You can click on a row to open a view with more detailed information about that specific sample.
- Gallery View: The Gallery View provides a more intuitive way to browse your actual data. For tabular data, the table displays the features used by your model, images are displayed as a grid of thumbnails, and text data is shown as a series of cards. You can adjust the size of the images in the gallery to see more or fewer items at a glance.
Filtering and Sorting¶
To help you find the data you're interested in, the Data Explorer provides a rich set of filtering options. You can filter your data by:
- Date and time range
- Batch index
- Sample ID
- Segments
You can also sort the data by any property that is not a list. Filters are automatically set to the latest data batch available in your task, and can be reset or cleared as needed.
Detailed Sample Exploration¶
When you click on a sample, a modal pops up, showing you all the available details for that sample. This includes:
- Tabular Data: A detailed view of all the features.
- Images: The full-resolution image. For object detection and semantic segmentation tasks, this also includes overlays on top of the prediction, target, or both.
- Text: The full text, and in the case of RAG (Retrieval-Augmented Generation) tasks, the user input, retrieved chunks, and model output. Any markdown syntax in the text is automatically rendered for better visual clarity.