🎛 UI Components
If you want to follow these examples hands-on, you can use the Spotlight example instances fired up in the 🚀 Getting Started section.
Examples will be annotated with the corresponding datasets name. You can use the tabs below for a quick reminder on how to load those datasets into Spotlight.
- mnist
- fsd50k
curl https://renumics.com/docs/data/mnist/mnist-tiny.csv -o mnist-tiny.csv
spotlight mnist-tiny.csv --dtype image=Image --dtype embedding=Embedding --dtype label=Category
curl https://spotlight.renumics.com/data/fsd50k/fsd50k-tiny.csv -o fsd50k-tiny.csv
spotlight fsd50k-tiny.csv --dtype window=Window --dtype embedding=Embedding --dtype audio=Audio
Spotlight UI provides a range of components that offer different perspectives on your data, allowing you to adapt the UI to your needs. You can add, remove, and rearrange these components as you wish to create a customized view of your data.
Once you have identified a component layout that works well for you, you can save it as a layout for future use. This way, you can quickly load your preferred layout whenever you need it, without having to reconfigure the UI each time.
Some of them like the Inspector and the Data Table help you in analyzing and understanding individual datapoints while others like the Similarity Map, Scatter Plot and the Histogram help you in understanding the overall distribution of your data.
The Filter Bar is a special component that allows you to filter your data based on the values of your features.
Filtering and selecting datapoints can affect how data is presented in the other components and therefore can greatly help you in analyzing and reasoning about your data.
Data Table
The data table often is the primary view on the data. In addition to several options that control which data is displayed, the table view also allows to edit datapoints. This includes the creation of new columns.
Inspector
The inspector lets you examine individual data points in depth by providing multiple different views for many of the data types supported by Spotlight. Including images, audio, 3D meshes and more.
Similarity map
The similarity map is a core element of most data-centric AI workflows. It allows for to map a given vector to a scatter plot by using a dimensionality reduction via UMAP or PCA. Different normalization options are available to handle metadata and embeddings. Additionally, dots on the scatter plot can be colored and sized.
Scatter plot
The scatter plot is typically useful as a supporting view to determine correlation between metadata information. Several aspects of the view can be customized dot size and color.
Histogram
The histogram view can be stacked to provide inside into data segments over two different dimensions.