AGri: Adaptive Thumbnails For Grid-based Visualizations

Keywords

Grid-based Visualization, Thumbnails, Spatiotemporal Data, Video Visualization

Abstract

This work introduces AGri (Adaptive Thumbnails for Grid-based Visualizations), a method for dynamically adjusting thumbnails of spatiotemporal data—such as videos—to varying screen footprints in grid-based layouts. AGri aims to maximize thumbnail expressiveness, which quantifies how well similarity relationships among data members (e.g., video frames) are preserved. Thumbnails are generated via cropping, with crop windows optimized based on cumulative salience images. By modeling the trade-off between expressiveness and footprint size, AGri defines a curve—the AGri curve—representing Pareto-optimal visual representations. This curve enables dynamic selection of thumbnails suited to different grid sizes and resolutions. The approach is demonstrated on two datasets: a spatiotemporal ensemble from scientific experiments and an animated short film.