Predicting Failures of Vision Systems



Computer vision systems today fail frequently. They also fail abruptly without warning or explanation. Alleviating the former has been the primary focus of the community. In this work, we hope to draw the community’s attention to the latter, which is arguably equally problematic for real applications. We promote two metrics to evaluate failure prediction. We show that a surprisingly straightforward and general approach, that we call Alert, can predict the likely accuracy (or failure) of a variety of computer vision systems – semantic segmentation, vanishing point and camera parameter estimation, and image memorability prediction – on individual input images. We also explore attribute prediction, where classifiers are typically meant to generalize to new unseen categories. We show that Alert can be useful in predicting failures of this transfer. Finally, we leverage Alert to improve the performance of a downstream application of attribute prediction: zero-shot learning. We show that Alert can outperform several strong baselines for zero-shot learning on four datasets.


P. Zhang, J. Wang, A. FarhadiM. Hebert and D. Parikh
Predicting Failures of Vision Systems
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014.
[Supp. material]


Mid-Atlantic Computer Vision (MACV) 2014 poster presentation [pdf]


Indoor scene images [Heduau et al. 2009] [data]
Bedroom and living room images [Satkin et al. 2012] [data]
Image memorability data [Isola et at. 2011] [data]
Animals with Attributes (AWA) [Lampert et al. 2009] [data]
UIUC [Farhadi et al. 2009] [aYahoo][aPascal]