Context Models and Out-of-context Objects

Myung Jin Choi Antonio Torralba Alan S. Willsky

Massachusetts Institute of Technology

 

Abstract

The context of an image encapsulates rich information about how natural scenes and objects are related to each other. Such contextual information has the potential to enable a coherent understanding of natural scenes and images. However, context models have been evaluated mostly based on the improvement of object recognition performance even though it is only one of many ways to exploit contextual information. In this paper, we present a new scene understanding problem for evaluating and applying context models. We are interested in finding scenes and objects that are “out-of-context”. Detecting “out-of-context” objects and scenes is challenging because context violations can be detected only if the relationships between objects are carefully and precisely modeled. To address this problem, we evaluate different sources of context information, and present a graphical model that combines these sources. We show that physical support relationships between objects can provide useful contextual information for both object recognition and out-of-context detection.

Publication

"Context Models and Out-of-context Objects"
Myung Jin Choi, Antonio Torralba, and Alan S. Willsky
To appear in Pattern Recognition Letters, 2012.

Downloads

Code MATLAB implementation of our support context model

Download supportContext.tar.gz (26MB). This tarball contains MATLAB scripts and our out-of-context dataset. To run object recognition on the SUN 09 dataset, you need to download the the data file below.

SUN 09 Dataset SUN 09 dataset and baseline detector outputs stored as MATLAB files (.mat)

Download datasetMat.tar (159MB). This file does not include images, so if you would like to display the detection results on images, you need to download sun09.tar (5.2GB) as well.

Latent-tree Learing Package Required only for training and not for testing our model

Download latentTree.tar (614MB). See Learning Latent Tree Graphical Models for more details.

Last update: December 18, 2011.