 |
    |
Undoing the Damage of Dataset Bias
Aditya Khosla1, Tinghui Zhou2, Tomasz Malisiewicz1, Alexei A. Efros2, and Antonio Torralba1
1Massachusetts Institute of Technology, USA
khosla@csail.mit.edu
tomasz@csail.mit.edu
torralba@csail.mit.edu
2Carnegie Mellon University, USA
tinghuiz@cs.cmu.edu
efros@cs.cmu.edu
Abstract. The presence of bias in existing object recognition datasets is now well-known in the computer vision community. While it remains in question whether creating an unbiased dataset is possible given limited resources, in this work we propose a discriminative framework that directly exploits dataset bias during training. In particular, our model learns two sets of weights: (1) bias vectors associated with each individual dataset, and (2) visual world weights that are common to all datasets, which are learned by undoing the associated bias from each dataset. The visual world weights are expected to be our best possible approximation to the object model trained on an unbiased dataset, and thus tend to have good generalization ability. We demonstrate the effectiveness of our model by applying the learned weights to a novel, unseen dataset, and report superior results for both classification and detection tasks compared to a classical SVM that does not account for the presence of bias. Overall, we find that it is beneficial to explicitly account for bias when combining multiple datasets.
LNCS 7572, p. 158 ff.
lncs@springer.com
© Springer-Verlag Berlin Heidelberg 2012