Abnormality Detection in Images

Babak Saleh, Ali Farhadi, Jacob Feldman, Ahmed Elgammal
Rutgers, The State University of New Jersey
University of Washington

Dataset Snapshot

Abstract

When describing images, humans tend not to talk about the obvious, but rather mention what they find interesting. We argue that abnormalities and deviations from typicalities are among the most important components that form what is worth mentioning. In this project we introduce the abnormality detection as a recognition problem and show how to model typicalities and, consequently, meaningful deviations from prototypical properties of categories. Our model can recognize abnormalities and report the main reasons of any recognized abnormality. We also show that abnormality predictions can help image categorization. We introduce the Abnormality Object Dataset and show interesting results on how to reason about abnormalities.

Publications


IJCAI 2016
AAAI 2016
ECCV 2014 Workshop on Parts and Attributes
CVPR 2013 paper(3MB)
CVPR 2013 supplementary materials(30MB)
Technical Report on Dataset and Human Subject Experiment

Abnormal Objects Dataset

Download AAAI 2016 paper dataset
Download AAAI 2016 paper dataset -- Dropbox link
Download Abnormal Objects Dataset (35MB)
Download Abnormal Objects Dataset - cropped images(117MB)
This dataset contains 6 object categories similar to object categories in Pascal VOC that are suitable for studying the abnormalities stemming from objects. We are developing the dataset to have images for all the classes in PASCAL (in total 20 classes). The current version of dataset contains:
  • 1001 Images
  • Full Images(Annotation including Bounding Box is available upon request)
  • 6 Object category
  • Objects:
  • Aeroplane, Boat, Car, Motorbike, Sofa, Chair
    Dataset Snapshot