Abstract

Determining dense semantic correspondences across objects and scenes is a difficult problem that underpins many higher-level computer vision algorithms. Unlike canonical dense correspondence problems which con-sider images that are spatially or temporally adjacent, semantic correspondence is characterized by images that share similar high-level structures whose exact ap-pearance and geometry may differ. Motivated by object recognition literature and re-cent work on rapidly estimating linear classifiers, we treat semantic correspondence as a constrained detec-tion problem, where an exemplar LDA classifier is learned for each pixel. LDA classifiers have two dis-tinct benefits: (i) they exhibit higher average precision than similarity metrics typically used in correspondence problems, and (ii) unlike exemplar SVM, can output globally interpretable posterior probabilities without cal-ibration, whilst also being significantly faster to train. We pose the correspondence problem as a graphi-cal model, where the unary potentials are computed via convolution with the set of exemplar classifiers, and the joint potentials enforce smoothly varying correspon-dence assignment. 1.