Sample Visualization

This page provides examples of the different visualizers supported in the platform

Intro and shared properties of the visualizers

In this page we present examples for the different visualizers, and how they are viewed in the platform. To better understand how to add these visualization via your code integration, refer to the visualizer functions.

The top bar of visualizers: Identifying a sample and its properties

The image visualizer for the MNIST dataset. Hovering over a specific sample allows to zoom into the image (see top left magnifying glasses over sample). The top of each sample shows its sample ID and abbreviation of its respective data subset (train, val, test).

All visualizers in the sample review window share a common top bar that shows the following:

  • Sample ID, together with an abbreviation of the subset it belongs to (train, validation, test)

  • On Hover - a magnifying glass action that allows to zoom in and out

  • The selected Metadata tags that were added

The different visualization types in the Dashboard

Heatmaps Visualizer

Each visualized input get a second, attention overlayed visualizer. This overlays the shared features of a concept that is clustered within the platform over the image. It answers the question: "What makes these samples part of the same group?"

Each Input that we connect to a visualizer using the mapping would automatically be accompanied by a corresponding second visualizer in the sample review panel. This visualizer is a heatmap visualization that overlays the shared features of a concept that is clustered within the platform over the image. It answers the question: "What makes these samples part of the same group?".

Controlling the color saturation and opacity of the heatmaps is possible via the top contoller of the heatmap visualizer.

Each input visualizer named "x" would result in a second

Metadata Visualizer

The metadata visualizer. In the top controller the option to keep only selected columns, reset selection, or export the data into .csv are presented by the Collumns, Reset, and Export button respectively.

The metadata visualzier is a defaultive visualizer that shows the metadata for all selected samples. The top controller of this visualizer allows to control and export the current table:

  • The Collumns button - Allows column selection and filtering for the table, narrowing the metadata to a relevant subset of the properties.

  • The Reset button - Allows to reset previous columm filtering to show all.

  • The Export button - Exports the current table into a .csv

Image Visualizer

The image visualizer for the MNIST dataset. Hovering over a specific sample allows to zoom into the image (see top left magnifying glasses over sample)

The image visualizer supports a visualization of images within the platform. Hovering over an image shows the RGB values. It also presents an option to zoom in or out, by clicking the magnifying glass over the image.

Semantic Segmentation Visualizer

Semantic Segmentation visualizer. Allows a dynamic selection of which labels to plot in the mask, and a control over the opacity

This visualizer visualizes the semantic segmentation of an image. Different properties of these settings could be controlled:

  • Choosing which label would be shown in the mask

  • Controlling what is the mask opacity

Object Detection Visualizer

Presenting the Object Detection visualizer. This visualizer allows you to visualize bounding boxes (BBs) over an image. It allows a dynamic seleciton of what BBs are shown by confidence or label. It also allow to view any metadata that is attached to the bounding box.

This visualizer visualises an image with bounding boxes. Different properties of these settings could be controlled:

  • Choosing which bounding box labels would be presented

  • Showcasing all bounding box metadata or hiding it

  • Controlling what are the minimal and maximal confidence of the bounding boxes that are presented.

Horizontal Bar Visualizer

The horizontal bar visualizer. Used to visualize logits and probabilities

The horizontal bar is used to viasualize logits and probabilities.

Graph Visualizer

The Graph visualizer. jsed to visualize time-series

The Graph visualizer is used to viasualize time series.

Text Visualizer

The Text visualizer

The Text visualizer is used to visualize raw text.

Text Mask Visualizer

The text mask visualizer. Used to visualize token classification tasks (NER, etc.)

The Text mask visualizer is used to visualize a TextMask, for token classification tasks such as NER.

PreviousPopulation ExplorationNextPopulation exploration settings

Last updated

Was this helpful?