Abstract

Scene parsing is the problem of assigning a semantic label to every pixel in an image. Though an ambitious task, impressive advances have been made in recent years, in particular in scalable nonparametric techniques suitable for open-universe databases. This paper presents the CollageParsing algorithm for scalable nonparametric scene parsing. In contrast to common practice in recent nonparametric approaches, Col- lageParsing reasons about mid-level windows that are designed to cap- ture entire ob jects, instead of low-level superpixels that tend to fragment ob jects. On a standard benchmark consisting of outdoor scenes from the LabelMe database, CollageParsing achieves state-of-the-art nonparamet- ric scene parsing results with 7 to 11% higher average per-class accuracy than recent nonparametric approaches.