DATASET FINGERPRINTS
Exploring Image Collections Through Data Mining

Konstantinos Rematas Basura Fernando Frank Dellaert Tinne Tuytelaars

to appear in CVPR 2015, Boston, MA.




Abstract

As the amount of visual data increases, so does the need for summarization tools that can be used to explore large image collections and to quickly get familiar with their content. In this paper, we propose \emph{dataset fingerprints}, a new and powerful method based on data mining that extracts meaningful patterns from a set of images. The discovered patterns are compositions of discriminative mid-level features that co-occur in several images. Compared to earlier work, ours stands out because i) it's fully unsupervised, ii) discovered patterns cover large parts of the images,often corresponding to full objects or meaningful parts thereof, and iii) different patterns are connected based on co-occurrence, allowing a user to ``browse'' the images from one pattern to the next and to group patterns in a semantically meaningful manner.

Materials for download

Paper

Bibtex
@inproceedings{RematasCVPR15,
author = {Konstantinos Rematas and Basura Fernando and Frank Dellaert and Tinne Tuytelaars},
title = {Dataset Fingerprints: Exploring Image Collections Through Data Mining},
booktitle = {CVPR},
year = {2015},
}

Discovered Patterns

Below is a visualization for the VOC 2012 summary where each box is a pattern. Mouse over a box to see its support.

Pattern-enabled Browsing

In this demo we use the discovered patterns to navigate through the image collection. Click on a pattern (blue bubble) to see in which images it appears (orange bubble). By clicking an image you can see which patterns appear there. *Work still in progress*


This work was supported by the ERC grand Cognimund