Collection Space Navigator: An Interactive Visualization Interface for Multidimensional Datasets

TL;DR

The paper addresses the challenge of understanding high-dimensional spaces in large digital artifact collections by introducing the Collection Space Navigator (CSN), a browser-based interactive visualization interface. CSN combines a central Projection Area with an Object Panel and a Control Panel to enable simultaneous exploration of multiple 2D projections and multidimensional filters. Key contributions include a modular, open-source design with configurable projections (e.g., UMAP, t-SNE), interactive dimension and advanced text filters, and export capabilities for filtered data. Demonstrations across art, newsreels, and text-to-image prompts illustrate CSN's versatility for research, curation, and communication, with potential applicability to non-visual data.

Abstract

We introduce the Collection Space Navigator (CSN), a browser-based visualization tool to explore, research, and curate large collections of visual digital artifacts that are associated with multidimensional data, such as vector embeddings or tables of metadata. Media objects such as images are often encoded as numerical vectors, for e.g. based on metadata or using machine learning to embed image information. Yet, while such procedures are widespread for a range of applications, it remains a challenge to explore, analyze, and understand the resulting multidimensional spaces in a more comprehensive manner. Dimensionality reduction techniques such as t-SNE or UMAP often serve to project high-dimensional data into low dimensional visualizations, yet require interpretation themselves as the remaining dimensions are typically abstract. Here, the Collection Space Navigator provides a customizable interface that combines two-dimensional projections with a set of configurable multidimensional filters. As a result, the user is able to view and investigate collections, by zooming and scaling, by transforming between projections, by filtering dimensions via range sliders, and advanced text filters. Insights that are gained during the interaction can be fed back into the original data via ad hoc exports of filtered metadata and projections. This paper comes with a functional showcase demo using a large digitized collection of classical Western art. The Collection Space Navigator is open source. Users can reconfigure the interface to fit their own data and research needs, including projections and filter controls. The CSN is ready to serve a broad community.

Figures (4)

• Figure 1: The Collection Space Navigator (CSN). The central Projection Area displays a x-y scatter plot of images based on the selected projection (e.g. UMAP, t-SNE), with filtered images greyed out, and mouse-over highlight. The Object Panel (left) shows a larger Object Preview of the highlighted image, together with Object Info based on selected metadata; Object Appearance visualizes clusters (optional), sets the projection thumbnail size (zoomed-out and zoomed-in). The Control Panel (right) allows for selection of Data and Projections; custom interactive Dimension Filters and Advanced Filters facilitate dataset exploration, analysis, and understanding (see text); the filtered object metadata and current projection view can be downloaded via Export Filtered Data.
• Figure 2: Examples of various 2D projections and visualization features in the CSN tool. a) UMAP projection with large thumbnails, providing a comprehensive view of the image content; b) UMAP projection with medium size thumbnails and cluster colors of categorical data selected in Object Appearance; c) UMAP projection with medium size thumbnails and filtered-out objects in grey; d) UMAP projection with small thumbnails and cluster colors, providing a more compact representation; e) t-SNE projection, showing an alternative dimensionality reduction technique; f) Simple x/y plot, here showing PC1 over time for temporal analysis. The flexibility of the CSN tool provides the ability to effectively explore and compare data across different visualization methods via displaying multiple different 2D projections, flexible based on data import configuration, selectable in Data & Projections, combined with thumbnail sizes and cluster highlights as adjustable via the Object View Settings.
• Figure 3: Interactive Dimension Filters. Left: Unfiltered Dimension Filters, consisting of range sliders with interactive histograms above them, showing the distribution of all objects along the slider’s dimension with the bin of the currently selected object highlighted in red; Center: Reducing the range of one slider affects the distribution of all dimensions, reflected by the histograms; Right: Bin Mode functionality is activated by clicking on a histogram, allowing the user to activate one bin at a time, with the Projection Area displaying the corresponding objects within the active bin. These interactive Dimension Filter features allow in-depth exploration and visualization of multi-dimensional data distributions allowing users to gain a deeper understanding of the relationships between data points.
• Figure 4: Two distinct use cases of the Collection Space Navigator. Left: Historical newsreels with rich metadata; Right: Investigating text-to-image generation through Stable Diffusion with linguistic basic concepts as prompts. By using domain specific metadata filters and projections, the CSN tool allows for deeper exploration providing insight into visual and thematic patterns.