The Galaxy Browser demonstrates a novel searching interface using 2D similarity-based visualization of image frames.
An image collection originating from a high dimensional feature space is projected to 2D visualization space. This is done using ISOSNE, a non-linear projection which aims at preserving the relations between images as much as possible. When the 2D visualization space is defined, the collection is clustered into k groups using k-means clustering algorithm. After user interaction, a set of k images is displayed. This set gives an overview of the collection. To allow for sufficient visibility, the overlap among images is reduced.