1 Hit in 5.7 sec

Covering up bias in CelebA-like datasets with Markov blankets: A post-hoc cure for attribute prior avoidance [article]

Vinay Uday Prabhu, Dian Ang Yap, Alexander Wang, John Whaley
2019 arXiv   pre-print
Attribute prior avoidance entails subconscious or willful non-modeling of (meta)attributes that datasets are oft born with, such as the 40 semantic facial attributes associated with the CelebA and CelebA-HQ  ...  In this paper, we address this and propose a post-hoc solution that utilizes an Ising attribute prior learned in the attribute space and showcase its efficacy via qualitative experiments.  ...  In Section 2, we cover the estimation procedure for learning the Inter-Attribute Ising Prior (IAIP). In Section 3, we propose our post hoc bias-corrective framework.  ... 
arXiv:1907.12917v1 fatcat:rqwh2at4hzfmrkuu4zkixvcwbi